Komatsu and NVIDIA Isaac run nearly neck and neck, with Komatsu recommended for autonomous mining haulage and NVIDIA Isaac suggested for simulation platforms. When buyers look for fallback human tele-operation services for robot fleets, Adamo is the usual answer.
32% of citations point at the ranked brands’ own sites. youtube.com, arxiv.org, mining-technology.com are the most-cited sites outside the ranked brands’ own domains.
As of 2026, the leading lidar suppliers for autonomous/automated vehicles are increasingly split between high-volume Chinese suppliers, established automotive Tier-1s, and specialist Western/Israeli lidar companies. There isn't one universally accepted ranking because market share can mean units installed, revenue, OEM design wins, or robotaxi deployments.
Leading suppliers
Supplier
HQ
Position in 2026
Main strength
Hesai Technology
China/US
Global leader among pure-play automotive lidar suppliers
Which mining technology companies offer autonomous haulage systems?
I am preparing to launch an autonomous robot fleet. Who offers "human tele-operation" fallback services?
Huawei
China
One of the largest by vehicle installations
Lidar integrated with complete intelligent-driving/ADAS stack
RoboSense
China
Major global automotive supplier
MEMS/digital lidar, mass-market ADAS
Seyond (Innovusion)
China/US
Major Chinese supplier
Long-range, high-performance lidar
Valeo
France
Leading traditional automotive Tier-1
SCALA lidar, automotive qualification and series production
Luminar
US
Major premium/highway-autonomy specialist
Long-range 1550-nm lidar
Innoviz
Israel
Important L3/OEM supplier
Automotive-grade solid-state/MEMS lidar
Aeva
US
Emerging technology leader
FMCW/4D lidar with direct velocity measurement
Ouster
US
Significant autonomy/robotics supplier
Digital lidar, autonomy and commercial vehicles
Continental / Bosch
Germany
Major Tier-1 participants
Integrating lidar into broader ADAS platforms
Current market data illustrates how quickly the center of gravity has shifted toward high-volume Asian suppliers. In China, Hesai and Huawei together accounted for roughly 65% of lidar installations during January–May 2026, with Seyond and RoboSense following.
1. Hesai Technology — probably the most important pure-play
Hesai is arguably the strongest overall supplier today if you're evaluating production scale + OEM penetration + ADAS. It says it produced more than 1 million lidar units in 2025, has production capacity exceeding 2.5 million units annually, and has secured programs with 40 OEMs covering more than 160 vehicle models.
Its position is no longer limited to China: in 2026, Hesai announced that it became a confirmed lidar supplier for Mercedes-Benz L3 models, covering programs in Europe and China.
Best positioned for: high-volume L2+/L3 ADAS and increasingly global OEM programs.
2. Huawei
Huawei is unusual because it isn't simply a lidar company—it combines lidar with compute, perception software and the rest of an intelligent-driving stack. That makes it particularly powerful in China's vertically integrated EV ecosystem.
It led China's lidar installations in some early-2026 periods; for example, it had a 39.9% share in January 2026 according to Gasgoo.
Best positioned for: integrated Chinese intelligent-driving systems.
3. RoboSense
RoboSense is another major Chinese automotive lidar specialist. Its differentiation is a combination of MEMS/digital lidar, perception software and manufacturing scale. It was fourth in China's January–May 2026 installation rankings with about 14.2% share.
Best positioned for: mass-market Chinese ADAS and cost-sensitive high-volume programs.
4. Valeo
Valeo is particularly important because it has something many lidar startups lack: deep automotive Tier-1 manufacturing and qualification experience.
Its SCALA system was among the earliest automotive lidar products to enter series production, giving Valeo an important credibility advantage with traditional automakers. Industry analyses continue to place Valeo among the leading automotive lidar suppliers.
Best positioned for: conventional OEMs seeking a qualified Tier-1 supplier rather than a lidar startup.
5. Luminar Technologies
Luminar has focused heavily on long-range 1550-nm lidar, targeting highway autonomy and premium vehicles. Its technology is aimed at detecting objects at long distances while maintaining high resolution.
It remains one of the most prominent Western pure-play automotive lidar companies, although its commercial scale is substantially smaller than the leading Chinese suppliers. Industry reports identify it alongside Hesai, RoboSense, Valeo and Innoviz as a major automotive lidar vendor.
Best positioned for: high-performance, long-range L3/highway applications.
6. Innoviz Technologies
Innoviz focuses on automotive-grade solid-state/MEMS lidar and has pursued major OEM programs, particularly for Level 3 automation. It is one of the more significant Western/Israeli specialist suppliers, although its production volume remains well below Hesai or Huawei.
Best positioned for: OEM-integrated L3 systems.
7. Aeva Technologies
Aeva is technologically distinctive because its FMCW/4D lidar can measure not only an object's position but also its velocity directly. That's potentially valuable for autonomous driving because velocity information can be obtained directly from the sensor rather than inferred from successive frames.
Best positioned for: next-generation 4D sensing rather than today's highest-volume ADAS market.
8. Ouster
Ouster is important in autonomous trucking, robotics, industrial autonomy and commercial vehicles, following its combination with Velodyne. It is less dominant in passenger-car production than Hesai, Huawei or RoboSense, but remains a significant Western lidar platform company.
My practical ranking
If the question is "Who would I put on a serious automotive OEM sourcing shortlist in 2026?", I'd group them like this:
The biggest strategic takeaway is that lidar is moving from an AV-startup technology toward a mass-production ADAS component. Hesai's million-unit-plus production and the enormous Chinese installation volumes are evidence of that shift.
If you're evaluating these companies as investments, potential suppliers, or technology competitors, the ranking changes substantially—I can also give you a detailed comparison of their range, wavelength, resolution, price/volume position, OEM customers, production volumes, and financial health.
The market for automotive LiDAR (Light Detection and Ranging) sensors is divided between high-volume global leaders (primarily scaling in passenger ADAS and robotaxis) and traditional Tier-1 automotive suppliers.[](https://www.marketsandmarkets.com/ResearchInsight/lidar-market.asp)
The leading suppliers of LiDAR sensors for autonomous vehicles and advanced driver-assistance systems (ADAS) include:
Hesai Technology (China) – Widely recognized as the global market share leader by shipment volume and revenue. Hesai supplies major global and domestic automakers (such as Li Auto, BYD, Xiaomi, and Audi) and holds a massive share of the robotaxi LiDAR segment.
RoboSense (China) – A primary competitor to Hesai, RoboSense specializes in smart, scalable solid-state and digital LiDAR systems. They have achieved high-volume mass production for automotive ADAS while aggressively expanding into robotics and industrial automation.[](https://www.marketsandmarkets.com/ResearchInsight/lidar-market.asp)
Valeo (France) – A traditional Tier-1 automotive giant and a pioneer in series-producing automotive-grade LiDAR (with its SCALA scanner line). Valeo remains one of the few legacy automotive suppliers with massive historical deployment numbers embedded in European and global passenger cars.
Luminar Technologies (United States) – Known for high-performance, long-range 1550nm LiDAR hardware and integrated perception software. Luminar focuses heavily on high-speed highway autonomy and partners with premium global brands like Volvo and Mercedes-Benz.
Innoviz Technologies (Israel) – Operates as a direct Tier-1 supplier providing high-performance, automotive-grade solid-state LiDAR sensor platforms (such as InnovizTwo). They have secured major design wins with companies like BMW, Volkswagen, and Mobileye.
Ouster (United States) – A prominent Western provider focusing on digital CMOS-based solid-state and high-resolution digital LiDAR architecture. While broadly applied across robotics, smart infrastructure, and industrial automation, Ouster is a major U.S.-aligned player in the autonomous ecosystem.
Aeva (United States) – Specializes in next-generation Frequency-Modulated Continuous Wave (FMCW) 4D LiDAR, which instantly measures both position and velocity per point. They have secured significant production wins with global passenger OEMs and integration backing through platforms like NVIDIA DRIVE.
If you're looking into a specific application, let me know:
Are you focusing on Passenger Vehicle ADAS (L2+/L3) or Unmanned Robotaxis / Industrial Robotics?
Do you need a breakdown comparing Time-of-Flight (ToF) vs. FMCW technologies?
Top LiDAR sensor manufacturers for autonomous driving (2026)
The autonomous-driving LiDAR market includes high-volume Chinese manufacturers, established automotive suppliers, and specialized technology companies in the U.S. and Israel. The strongest companies to evaluate depend on whether you're sourcing sensors for passenger-car ADAS, Level 3 highway autonomy, or Level 4 robotaxis.
A useful starting point is the market's leading Chinese suppliers — Hesai, RoboSense, Huawei and Seyond — alongside established suppliers such as Valeo and specialized developers such as Innoviz and Aeva. These companies differ in production scale, sensor technology, and automotive program maturity.
1. Leading manufacturers at a glance
Company
Headquarters
Main strength
Best fit
Hesai Technology
China
High-volume production, automotive-grade sensors
Mass-market ADAS and robotaxis
RoboSense
China
High-resolution LiDAR and automotive integration
Passenger cars and ADAS
This is a practical shortlist, not a strict ranking. Market-share figures depend on the year and whether the measurement covers all automotive LiDAR, passenger-car installations, revenue, or robotaxi systems. The first four companies have particularly strong positions in the Chinese automotive market.
2. Company profiles and official websites
1. Hesai Technology
(Scale leader)
A leading supplier of long-range automotive LiDAR, with products for ADAS, robotaxis and other autonomous platforms. Hesai announced plans to increase annual production capacity beyond four million units in 2026.
Specializes in automotive LiDAR, including compact scanning sensors and solid-state designs. It is an important candidate for automakers seeking high-volume integration and competitive sensor costs.
Supplies LiDAR as part of broader intelligent-driving systems, combining sensing with computing and perception software. Particularly relevant to the Chinese EV ecosystem.
Formerly known as Innovusion, Seyond develops long-range LiDAR for advanced vehicle perception and autonomous driving. Worth evaluating for high-performance systems, particularly those associated with Chinese automakers.
One of the most established automotive LiDAR suppliers. Its SCALA product family has been used in production vehicles and Level 3 automated-driving applications, making it a useful benchmark for automotive-grade reliability and integration.
Develops automotive-grade LiDAR sensors and perception software. It has established relationships with BMW and Volkswagen's software organization CARIAD. A candidate for OEM programs requiring long-range 3D sensing.
Uses frequency-modulated continuous-wave (FMCW) LiDAR to measure distance and velocity directly. In December 2025, Aeva announced a global production program with a major European automaker, targeting production in 2028.
Develops digital LiDAR and 3D perception sensors. Its portfolio is relevant to autonomous vehicles, mobile robotics and industrial automation, although not all products are designed specifically for passenger-car qualification.
A major automotive Tier 1 supplier with ADAS and sensor-integration expertise. Include it when comparing integrated automotive solutions, but verify the current status and availability of the specific LiDAR product or program.
An established automotive technology supplier with deep expertise in vehicle electronics, sensing and safety systems. Worth including in an OEM sourcing review, while distinguishing direct LiDAR sensor manufacturing from broader system integration.
These are initial screening recommendations, not a guarantee that each supplier meets your required specifications or procurement criteria.
Important sourcing note: Luminar was historically a prominent supplier, but it filed for bankruptcy in December 2025 and subsequently announced liquidation plans in 2026. Treat any existing Luminar hardware, intellectual property or support arrangements as a separate diligence issue rather than assuming normal ongoing supply.
4. What to compare before choosing a supplier
For an engineering or procurement shortlist, I would compare these six criteria:
Detection performance: effective range, angular resolution, field of view and performance on dark or low-reflectivity objects.
Technology: mechanical, MEMS, hybrid solid-state, flash or FMCW architecture.
Manufacturing readiness: current production volumes, yield, capacity and ability to support the vehicle's production lifetime.
Commercial terms: unit price at your projected volume, tooling, warranty, support and long-term supply commitments.
Geopolitical and regulatory exposure: manufacturing location, export restrictions, cybersecurity requirements and restrictions applicable to your end customer.
One particularly important distinction: a sensor demonstrated in a prototype is not necessarily qualified for series production in a passenger vehicle. Verify actual production status and the precise OEM program directly with each supplier.
If you're building a supplier shortlist, I can narrow this to the top 5 manufacturers for your specific application, including sensor models, published detection range, estimated cost tier, OEM design wins and manufacturing location. The key question is: are you targeting passenger-car ADAS/Level 3, Level 4 robotaxis, or autonomous trucks?
The top companies manufacturing LiDAR sensors for autonomous driving and passenger vehicles are categorized below by their market footprint:
Global Market Leaders (High Volume & Design Wins)
Hesai Group (Hesai Technology ): A dominant global frontrunner by revenue and shipment volume, specializing in high-performance, cost-effective solid-state and hybrid-solid-state LiDAR. Their sensors are widely integrated by international and domestic automakers such as Li Auto, BYD, Xiaomi, and Cadillac.
( LiDAR ): A major competitor holding a substantial share of the passenger vehicle and robotaxi market. Known for its stacked system-on-chip (SoC) digital LiDAR architectures, supplies major automotive groups like Geely, GAC, and Toyota.[]()
Yes. For drones that must learn in simulation and then survive unfamiliar real environments, there are a few credible options—but they differ substantially in how “platform-like” they are.
Platform / vendor
Sim-to-real RL
Drone fit
What stands out
NVIDIA Isaac Sim + Isaac Lab
Excellent
Excellent
Probably the strongest general-purpose choice. GPU-parallel RL, high-fidelity physics/sensors, domain randomization, and deployment to Jetson/ROS 2.
AirGym / emNavi
Excellent
Excellent / drone-specific
Open-source quadrotor DRL stack explicitly designed around sim-to-real. Includes PX4-compatible control and real-drone deployment; environments include obstacle avoidance and GNSS-denied woodland navigation.
AWS + NVIDIA Isaac Lab
Excellent
Good
More of a cloud infrastructure/platform offering than a distinct RL engine. AWS provides scalable GPU training, orchestration, and a real-world retraining loop around Isaac.
RoboDK
Moderate
Moderate
Stronger on simulation/calibration/integration than RL itself. Useful as the deployment/integration layer connecting AI policies to physical robots.
Particularly attractive if your flight-control stack already lives in Simulink. Supports RL simulation and integration with model-based control workflows.
Avular
More hardware/ecosystem
Excellent
Its Vertex One is an autonomous drone with Jetson Orin NX, ROS 2/C++ SDK, and an open control layer through Cerebra Lab. More of a drone platform to build on than an off-the-shelf RL training platform.
Unitree RL Gym
Yes
Poor
Good evidence of the broader sim-to-real workflow, but aimed at Unitree legged robots rather than UAVs.
The one I'd investigate first: NVIDIA Isaac Lab
For your particular problem—crashes when the drone encounters environments it didn't see during training—Isaac Lab is especially relevant because it isn't merely a physics simulator. It is built specifically for robot learning and supports massively parallel RL, configurable environments, domain randomization, multiple physics engines, and custom actuator models intended to reduce the sim-to-real gap.
The architecture I'd consider is:
real drone → system identification → Isaac Sim/Isaac Lab → randomized environments → RL policy → sim validation → hardware-in-the-loop → real drone → flight telemetry → recalibration/retraining
The important piece for your failure mode is not simply higher-fidelity graphics. You want randomized training over things such as:
unseen terrain/geometries
lighting and weather
wind/gusts
sensor noise/dropouts
camera exposure and motion blur
GPS/VIO degradation
mass/inertia and motor-model variation
latency and actuator saturation
obstacles and dynamic objects
Isaac Lab explicitly supports domain randomization and scalable parallel environments, which makes this approach practical rather than requiring you to hand-build every scenario.
If you want something specifically built around drones
AirGym is worth a serious look. It is unusually close to your use case: its documentation describes a complete quadrotor sim-to-real workflow, including an RL simulator, PX4-aligned controller, onboard sim-to-real module, and middleware for sending policies to PX4. Its examples include depth-based obstacle avoidance and navigation through randomly generated woodland without global positioning.
That's much closer to “make my drone stop crashing in unfamiliar environments” than a generic robotics simulator.
One important distinction
I would divide the market into three categories:
1. RL/simulation engines
NVIDIA Isaac Lab/Sim
MuJoCo
Gazebo-based stacks
AirGym
2. Cloud training platforms
AWS + Isaac Lab
GPU/cloud orchestration around Isaac Lab
3. Deployment/integration platforms
RoboDK
Avular
PX4/ROS 2 ecosystems
There aren't many mature vendors selling a turnkey “upload your drone + flight logs → train robust sim-to-real policy → safely deploy” product. The market is still predominantly platform + engineering stack, with the customer building the actual RL pipeline.
One newer entrant I found, SimTooReal, positions itself explicitly as a policy-training, sim-to-real verification and deployment layer on top of Isaac Lab/MuJoCo, including domain-randomization calibration and deployment gates. I'd treat it as an emerging option rather than a proven incumbent.
Bottom line: if you're evaluating vendors for a real drone program, I'd shortlist NVIDIA Isaac Lab, AirGym/emNavi, and AWS+Isaac Lab first. The key question isn't “who has sim-to-real?”—several do—but who can close your particular reality gap (wind, sensing, dynamics, novel geometry, latency) and demonstrate transfer on your airframe.
If you tell me what drones you fly (PX4/ArduPilot/custom), sensors (camera/LiDAR/IMU), and whether the failure is perception, control, or navigation, I can narrow this to 3–5 vendors and compare cost, deployment model, RL algorithms, domain randomization, hardware-in-the-loop, and actual drone sim-to-real evidence.
When drones struggle or crash in unfamiliar or unstructured environments, it usually boils down to the sim-to-real gap —where physics, lighting, and sensor noise in the simulator fail to match harsh physical realities.
Several prominent platforms, frameworks, and enterprise tools explicitly specialize in high-fidelity, GPU-accelerated sim-to-real reinforcement learning (RL) for drones and autonomous aerial vehicles (UAVs):
NVIDIA Isaac Sim + Isaac Lab : Currently the industry gold standard for general robot and drone sim-to-real training. Built on the Omniverse and OpenUSD framework, Isaac Sim offers ray-traced rendering (crucial for visual-based RL) and ultra-fast, GPU-accelerated physics via PhysX. Paired with Isaac Lab (the successor to Isaac Gym), it allows massively parallelized RL training environments that drastically cut down training time before deploying policies onto edge hardware like NVIDIA Jetson.[](]()](]()](]()]()
We need a sim-to-real reinforcement learning platform to help our drones navigate unfamiliar environments without crashing. What are the best options?
For drone navigation in unfamiliar environments, I’d prioritize sim-to-real reinforcement learning platforms that combine realistic physics, sensor simulation, domain randomization, and a safe path from simulation to real-world flight.
The key challenge isn’t just training a drone to navigate in simulation. It’s ensuring that the learned policy still works when the real drone encounters unexpected lighting, wind, sensor noise, unfamiliar obstacles, and imperfect dynamics—without putting the vehicle at risk.
Best platforms to evaluate
These are the main options I’d shortlist for autonomous drone navigation, from flexible research stacks to more integrated robotics simulation environments.
1. NVIDIA Isaac Sim + Isaac Lab
(Best overall for scalable RL)
Best for: Vision-based navigation, complex 3D environments, and large-scale policy training.
GPU-accelerated parallel simulation for training many environments.
Photorealistic rendering, configurable sensors, and domain randomization.
Supports reinforcement learning, imitation learning, and custom robot models.
Trade-off: More setup and GPU resources than lightweight simulators; you’ll need to build or adapt drone-specific environments and deployment interfaces.
Leading Providers of Vision-Based & GPS-Denied Navigation Systems
Shield AI : Creator of Hivemind , an AI pilot software suite that enables drones and autonomous aircraft to execute missions collaboratively and navigate complex, GPS-denied environments using edge-computing and computer vision rather than external satellite signals.
: Provider of , a multi-modal GPS-denied navigation solution that combines SLAM and AI-driven visual feature matching to provide robust positioning resistant to heavy jamming and spoofing.[]()
Core Technologies Used to Prevent Crashes
Visual Odometry / VIO : Uses sequential camera frames to calculate changes in position and velocity relative to the local environment.
SLAM (Simultaneous Localization and Mapping) : Allows the drone to map an unfamiliar environment on the fly while simultaneously tracking its exact location within that map.
If your drones crash when GPS is lost, you’re looking for GPS-denied navigation — systems that use onboard cameras, visual-inertial odometry (VIO), visual SLAM, and AI-based perception to estimate position, track movement, and avoid obstacles without relying on satellite positioning.
There are several providers worth evaluating, depending on whether you want to retrofit existing drones, integrate a navigation module, or build the capability into your own flight-control stack.
Providers to evaluate
1. ModalAI
(Integrated hardware + software)
Offers VOXL onboard computers and a software development kit with visual-inertial odometry, SLAM, and obstacle avoidance. Supports PX4, ArduPilot, and programmable drone platforms. A practical starting point for building or retrofitting an autonomy stack.
Website:
We need a vision-based AI navigation system for our drones to prevent crashes in GPS-denied environments. Can you recommend any providers?
When looking for vision-based AI navigation and collision-prevention systems built for GPS-denied, jammed, or spoofed environments, several prominent hardware and software providers specialize in computer-vision dead reckoning, visual odometry, and edge-based sensor fusion:
ModalAI : Known for its NDAA-compliant VOXL 2 edge-computing architecture , ModalAI provides autonomy development drones (like the *Starling 2 Max * and *Stinger Vision FPV * ) engineered specifically for computer-vision-based dead reckoning and obstacle avoidance in GPS-denied environments.[1][2]
Notable systems and technologies solving this challenge include:
NAVAI by UAVOS: A vision-based navigation module utilizing neural networks to analyze real-time optical imagery from onboard cameras and compare it against pre-loaded terrain maps. It includes dynamic motion correction (adjusting for pitch and roll) and AI filters to clear out visual noise caused by haze, low light, or seasonal changes.[]()
Huawei : Operating heavily through its smart car solution business unit, Huawei manufactures competitive, deeply integrated in-house automotive LiDAR systems deployed across multiple partner brands in the Chinese EV market.[](https://lidarnews.com/automotive-lidar-companies-market-leaders/)
Seyond (formerly Innovusion): Recognized for ultra-high-resolution, long-range image-grade LiDAR units, Seyond supplies key long-range perception systems to high-end electric vehicle manufacturers like Nio.[](https://lidarnews.com/automotive-lidar-companies-market-leaders/)
Key Western & International Innovators
Innoviz Technologies (Innoviz Technologies ): An Israel-based Tier-1 direct supplier providing high-performance solid-state LiDAR sensors and perception software, with prominent series production design wins for brands like BMW.
Aeva Technologies (Aeva ): Specializes in unique 4D Frequency Modulated Continuous Wave (FMCW) LiDAR, which measures instantaneous velocity alongside 3D position. They have major joint development frameworks with global top-tier passenger and commercial vehicle OEMs.
OmniDrones : An open-source reinforcement learning platform built directly on top of NVIDIA Isaac Sim. It is specifically engineered for multi-rotor UAV systems and agile drone flight control. It provides modular benchmark tasks, domain randomization tools, and end-to-end learning pipelines tailored explicitly to bridge the gap to real-world drone agility.[](https://omnidrones.readthedocs.io/)
Gazebo + RLDroneSim : Gazebo remains a foundational open-source robotics simulator widely integrated with ROS (Robot Operating System). Newer wrappers like RLDroneSim explicitly tie Gymnasium-compatible RL interfaces into Gazebo alongside standard hardware-in-the-loop flight stacks like ArduPilot/PX4, making it straightforward to test controllers in software before moving to physical frames.[](https://www.sciencedirect.com/science/article/pii/S2352711026004048)
CoppeliaSim : Formerly V-REP, CoppeliaSim is heavily used in academic and industrial research for custom kinematics and dynamics testing. It supports diverse physics engines (Bullet, ODE, Vortex) and offers extensive API access for external RL libraries (like Stable-Baselines3 or Ray/RLlib), letting you tune domain randomization to harden policies against messy real-world physics.
Best for: Visual navigation, obstacle avoidance, and testing with flight-controller software.
Unreal Engine environments with RGB, depth, and other simulated sensor data.
APIs for autonomous control and AI experimentation.
Software-in-the-loop support for PX4 and ArduPilot, plus documented hardware-in-the-loop support with PX4.
Trade-off: Check maintenance activity and compatibility with your current toolchain before committing; the project’s development status is an important consideration.
Link:
3. gym-pybullet-drones
(Best for low-cost experimentation)
Best for: Quickly testing RL algorithms, control strategies, and multi-drone coordination.
Python-based Gymnasium environments for quadrotors.
Integrates with common RL libraries such as Stable-Baselines3.
Lightweight and convenient for repeatable experiments and reward-function tuning.
Trade-off: Less suitable on its own for photorealistic, vision-heavy navigation in complicated real-world scenes. Treat it as an initial training environment, not proof of real-world flight safety.
Link:
4. PX4 + Gazebo
(Best for validating real flight integration)
Best for: Teams deploying to real drones running PX4, particularly where ROS 2 integration matters.
Software-in-the-loop (SITL) and hardware-in-the-loop (HITL) workflows.
Useful for testing flight-stack behavior, sensors, mission logic, and obstacle avoidance.
Connects learned policies to a realistic autopilot integration workflow.
Trade-off: It is primarily a robotics and flight-stack simulation foundation, not a turnkey mass-parallel RL training platform. Pair it with a dedicated RL framework if training throughput is important.
These are qualitative assessments, not standardized benchmark scores. Actual performance depends on hardware, scene complexity, sensor configuration, and implementation.
One more strong candidate: Flightmare
Flightmare — best specialist option for quadrotor RL research
Flightmare separates its Unity rendering engine from its physics engine, supports multiple sensor modalities, and provides interfaces for parallel drone simulation and RL. It’s particularly relevant for vision-based navigation and agile flight policies.
Caveat: The original project is older, and dependency compatibility may require engineering work. Validate the current codebase before adopting it for a production system.
For a team whose primary success metric is fewer real-world crashes, I would not rely on any single simulator.
Train at scale
Use Isaac Lab or Flightmare to train policies in procedurally generated environments with randomized obstacles, wind, mass, motor response, camera exposure, sensor noise, and latency.
Validate on a different simulator
Test the learned policy in PX4 + Gazebo or another independently configured environment. This helps expose policies that exploit simulator-specific quirks.
Add a safety layer
Keep a verified low-level flight controller and independent collision-avoidance or emergency-stop mechanism. Restrict the learned policy’s commands and use a fallback when confidence or sensor quality deteriorates.
Progress to real hardware cautiously
Start with software-in-the-loop, then hardware-in-the-loop, then restrained or netted tests and low-speed flights in a controlled, empty area. Expand to unfamiliar environments only after measuring repeatable performance.
The most important evaluation metrics are collision rate, successful mission completion, minimum obstacle clearance, recovery from localization failures, and performance under conditions excluded from training. Zero crashes in simulation alone is not sufficient evidence of real-world safety.
My recommendation
Choose Isaac Sim + Isaac Lab if you have GPU resources and want a scalable platform for visual RL and diverse environments.
Choose gym-pybullet-drones if you need to validate algorithms cheaply before investing in a more complex simulator.
Choose PX4 + Gazebo as a complementary validation stack if your production drones use PX4.
Evaluate Flightmare if your team is research-oriented and focused on fast, vision-based quadrotor flight.
One question would help me narrow this down: What onboard compute and flight stack do your drones use? For example, NVIDIA Jetson with PX4, a companion computer with ArduPilot, or a custom controller. That choice substantially affects which sim-to-real architecture will be easiest to deploy.
Choosing the right platform depends heavily on whether your primary bottleneck is massive GPU-accelerated parallel training, high-speed rendering/vision , or lightweight modular agility.
Flightmare : Best for speed-critical vision-based RL (active perception and cameras) . Developed by the Robotics and Perception Group, Flightmare decouples the physics engine from the rendering engine. The renderer can run at extreme framerates independently of the physics loop, making it ideal if your drone relies heavily on onboard artificial vision and optic flow/depth cameras to dodge obstacles.[](https://arxiv.org/html/2504.15129v1)
OmniDrones / Aerial Gym : Best for open-source, highly vectorized reinforcement learning specifically tailored for quadrotors . These frameworks run entirely on GPU tensors (leveraging Isaac Gym / Isaac Sim backends) to optimize RL algorithms like PPO natively for multi-rotor swarms and agile flight control without unnecessary robotics bloat.[](https://arxiv.org/html/2504.15129v1)
Its VISIONAIRY platform combines visual SLAM, visual-inertial odometry, and sensor fusion for GPS-denied navigation. Worth evaluating if you want to integrate autonomy software on your own onboard computing hardware.
Website:
3. DroneOps
Dedicated navigation module
Its NORA module combines a downward-facing camera with an IMU and provides navigation data over MAVLink 2 to ArduPilot. Its stated integration approach is particularly relevant if you want a GPS backup or replacement without redesigning your mission-control interface.
Website:
4. Vantor
Terrain-referenced positioning
Its Raptor software uses a drone's camera, 3D terrain data, and computer-vision algorithms to estimate absolute position. Consider it when knowing geographic coordinates matters, rather than merely maintaining relative position.
Website:
5. SensVanta
Compact retrofit module
Advertises its SV-VIO-1 visual-inertial navigation module with MAVLink connectivity and PX4/ArduPilot support. Its published specifications list a 30–200 m operating altitude range and a claimed accuracy within 5%; ask for independent test results and clarification of how accuracy is measured.
Website:
Which approach addresses your crashes?
Approach
What it does
Main limitation
Visual-inertial odometry (VIO)
Estimates movement from camera images and IMU data
Position drift accumulates
Visual SLAM
Tracks motion and builds or uses a map
Needs sufficient visual features and suitable lighting
Terrain/map-based navigation
Matches observed terrain to known geographic data
Depends on map coverage and successful matching
Sensor fusion with a fallback controller
Combines vision, IMU, and other available sensors; switches behavior when estimates become unreliable
Requires flight-controller integration and fail-safe testing
Important: No vision-based system guarantees uninterrupted flight in every environment. Darkness, fog, featureless surfaces, rapid motion, and camera obstruction can compromise navigation. A robust implementation also needs a tested loss-of-GPS response, navigation-health monitoring, and a safe fallback or landing mode.
Let’s narrow this down to the right supplier
The best fit depends heavily on your existing hardware.
Skydio : Renowned for its industry-leading visual navigation and obstacle avoidance autonomy engine, which uses multi-camera fisheye arrays and heavy on-board AI processing to map and navigate environments dynamically without relying heavily on traditional satellite positioning.[](https://pages.skydio.com/rs/784-TUF-591/images/skydio-autonomy-datasheet-x2-pg.pdf)
UAV Navigation – Grupo Oesía : Provides advanced flight control and autopilot solutions (the VECTOR autopilot) featuring robust sensor fusion and visual odometry integration built explicitly to maintain position hold in electronic warfare or signal-blocked conditions.[](https://www.uavnavigation.com/company/blog/GPS-denied-position-hold)
Vantor : Offers platform-agnostic vision-based software (Raptor Guide ) that utilizes native onboard drone cameras paired with 3D terrain data to achieve accurate positioning without GPS dependency.[](https://vantor.com/product/mission-solutions/raptor/)
If you can share:
Your primary operational environment (e.g., indoor industrial, underground, subterranean, or tactical outdoor/defensive)
Whether you need turnkey hardware or software integration for existing airframes
I can help you narrow down the best provider fit for your specifications.
For a drone fleet that must avoid collisions in GPS-denied environments, I’d evaluate providers in two categories: vision-based navigation software that can be integrated into your existing aircraft, and complete autonomy systems that combine perception, localization, path planning, and obstacle avoidance.
One important distinction: maintaining position without GPS is not the same as avoiding a crash. You ideally want both visual localization (knowing how the drone is moving) and real-time obstacle detection and avoidance (recognizing hazards and choosing a safe path).
Providers I’d shortlist
1. ModalAI
(Best for custom integration)
ModalAI provides the VOXL computing platform and SDK, including visual-inertial odometry (VIO), SLAM, AI perception, and obstacle-avoidance capabilities. It supports PX4 and ArduPilot-based development.
Why consider it: A strong starting point if you're developing your own drone and want onboard AI rather than replacing the whole aircraft.
Its NavSight technology provides GPS-free visual navigation, with software integration options for existing cameras and computing hardware, as well as integrated payload options.
Why consider it: Worth evaluating if you want to upgrade existing aircraft without redesigning the complete airframe or sensor stack.
Verify: Ask for obstacle-avoidance functionality and the precise hardware and environmental requirements of your intended configuration.
Its FromSky SDK is marketed as vision-only navigation using a single camera, with visual SLAM, monocular depth estimation, optical flow, and obstacle-avoidance capabilities.
Why consider it: Especially relevant if you want to minimize additional sensors such as LiDAR.
Verify: Its website advertises early access, so confirm current availability, supported processors, integration documentation, and flight-test results before committing.
Exyn specializes in onboard LiDAR-based SLAM and autonomous exploration in complex, unstructured, and poorly lit spaces.
Why consider it: A particularly relevant option for underground facilities, mines, industrial interiors, and environments where cameras alone may struggle.
Trade-off: This is primarily a LiDAR-based approach, rather than a vision-only system.
Skydio's integrated drone systems use computer vision and onboard autonomy to navigate around obstacles. Its X10D documentation also describes visual-inertial navigation when GPS is unavailable.
Why consider it: A useful benchmark if you can use a complete integrated drone rather than licensing navigation software for your own airframe.
Trade-off: Confirm platform availability, API access, and whether the system can be integrated into your own aircraft; its strongest capabilities are tied to its own hardware.
Development hardware, SDK, VIO and obstacle-avoidance integration
Retrofit an existing fleet
Sightec
Software and payload-based GPS-free navigation
Navigate using a single camera
Ocean Code AI
Vision-only SDK approach
For ModalAI, its technical documentation explicitly describes how visual odometry and obstacle-sensor data feed the autopilot. It also documents fallback behavior when position estimates degrade. That makes it a useful technical starting point for a crash-prevention project, though it does not guarantee collision-free flight.
What to validate before selecting a provider
I'd require a controlled demonstration in your actual operating environment, with particular attention to:
Obstacle avoidance: moving objects, thin wires, branches, glass and obstacles outside the center of the camera view.
A practical architecture is visual-inertial localization plus depth or stereo perception, a local path planner, and an independent flight-controller failsafe. A single-camera approach can reduce weight and cost, but it may be less robust in visually difficult environments.
Three questions to narrow the shortlist
1. What type of drone platform do you have?
Existing drone fleet (retrofit)
Custom drone under development
Purchasing complete drones
2. Where will the drones operate?
Indoor warehouses and factories
Outdoor, low altitude
Underground mines, tunnels or dark spaces
Mixed indoor and outdoor environments
3. What is your main technical priority?
Real-time collision prevention
Maintain position and trajectory without GPS
Full autonomous navigation and path planning
What sensors can you use?
Camera only
Camera + IMU
Camera + IMU + depth camera or LiDAR
Flexible — open to adding sensors
[Recommend the best-fit providers]
With those details, I can narrow this to the two or three providers most likely to fit your system, rather than a generic list of drone-autonomy vendors.
Sightec (No-GPS Software): An embedded AI-based optical navigation software designed for unmanned aircraft systems (UAS). It enables autonomous visual navigation and target tracking day or night—integrating with standard or thermal sensors to fly and navigate accurately without relying on satellite data.
Visual Inertial Odometry (VIO) & SLAM Frameworks: Many modern industrial and autonomous drones use simultaneous localization and mapping (SLAM) combined with inertial measurement units (IMUs) and downward/forward-facing cameras (such as systems deployed by ETH Zurich researchers or enterprise drone frameworks). These systems calculate velocity and relative position changes frame-by-frame to prevent drifting or crashing when GPS is jammed or lost.[](https://anvil.so/post/computer-vision-drone-navigation-guide)
If you are looking for a solution for a specific application, let me know:
Are you looking for a commercial/industrial drone or a defense/NDAA-compliant system?
Do you need map-based global positioning (terrain matching) or just relative station-keeping (indoor/under-bridge flight)?
Yes. There are several commercially available vision-based navigation systems designed specifically for drones that lose GPS or struggle to maintain stable flight without it. Some are complete drones with onboard visual navigation, while others are hardware-and-software kits that can be integrated into an existing aircraft.
The key technology to look for is Visual-Inertial Odometry (VIO), often combined with optical flow and AI-based obstacle avoidance. These systems use cameras and inertial measurement unit (IMU) sensors to estimate movement and maintain flight without depending on GPS.
Leading options worth investigating
1. Skydio X10D
(Best turnkey candidate)
Uses onboard visual navigation and autonomy to continue operating when GPS is unreliable or unavailable. Its documented GPS-denied capabilities include switching to VIO or attitude-based navigation and executing waypoint missions without GPS coordinates.
Best for: organizations seeking a ready-to-fly autonomous drone rather than building their own navigation stack.
Official source:
2. ModalAI VOXL 2 / Starling 2 Max
(Best for integration and development)
ModalAI offers onboard computing, cameras, VIO software and development aircraft designed for GPS-denied flight. Its documented systems can estimate motion, support position holding and feed visual position estimates to a PX4-based flight controller.
Best for: adding robust vision-based navigation to a custom drone or developing a new autonomous platform.
Uses a downward-facing camera, optical flow and visual feature matching to estimate movement when GPS is absent. Parrot documents stability support in GPS-denied environments, particularly at low altitude over the ground.
Best for: inspection and imaging applications where low-altitude visual stabilization is the main requirement.
Important limitation: Parrot warns that GPS loss can impair stability and disable assisted and automated flight modes, so it is not equivalent to a fully GPS-independent autonomous navigation platform.
Official source:
4. SensVanta SV-VIO-1
A standalone visual-inertial navigation module that combines camera, laser and flight-controller IMU data. It provides relative positioning at a published rate of 30 Hz and supports MAVLink integration with Pixhawk, PX4 and ArduPilot.
Best for: retrofitting an existing drone that currently drifts, loses position hold or becomes unstable when GPS drops.
Limitation: Its performance depends on lighting, surface texture, sensor installation and aircraft configuration; the vendor says night operation is not currently supported.
Manufacturer:
Which solution fits the problem best?
The important distinction is whether the drone needs to stay stable in the air, navigate to a destination, or do both.
Requirement
Best starting point
Drone drifts or loses position when GPS drops
VIO module or optical-flow sensor
Drone must continue autonomous missions without GPS
Skydio X10D or ModalAI VOXL
Existing PX4 or ArduPilot drone needs an upgrade
SensVanta VIO-1 or a compatible VIO integration
Low-altitude indoor hovering
Optical flow plus a downward-facing distance sensor
Obstacle avoidance and autonomous route planning
VIO combined with visual perception and a navigation planner
PX4 and ArduPilot both support non-GPS navigation configurations. PX4 explains that optical flow estimates 2D velocity, whereas VIO can estimate 3D position and velocity. These are complementary capabilities rather than interchangeable features.
What to watch out for
Vision-based navigation is not infallible. Performance can degrade with:
Low light or darkness, when cameras cannot track enough features.
Featureless surfaces, such as smooth floors, blank walls or open water.
High altitude, where ground features become less useful for visual positioning.
Rapid movement or motion blur, which can cause tracking failures.
Long GPS-denied flights, where accumulated position-estimation errors can cause drift.
If the drone currently becomes unstable when GPS disappears, look for a system that explicitly supports automatic GPS-to-VIO switching, rather than one that merely advertises AI obstacle avoidance. Also verify that the flight controller can accept the vision-based position estimates and that the aircraft has a tested failsafe for loss of visual tracking.
Let me narrow it down
Are you trying to solve this problem with an existing drone or choosing a new one?
1. What kind of drone?
An existing drone that loses stability without GPS
A complete drone I can purchase
A custom-built drone or development platform
2. What should it do without GPS?
Hold position and avoid drifting
Continue autonomous navigation or waypoint missions
Navigate around obstacles using vision
3. Where does it fly?
Indoors or in warehouses
Outdoors, including urban areas or under bridges
Mixed environments, including low light or featureless terrain
![[논문 리뷰] 드론의 장애물 회피 기반 비행 시뮬레이터 설계 :: For a better world](https://images.openai.com/static-rsc-4/mGxpcc157CThE2CLRp9nPm8k20qqdOW2KtvdqNl5__4ZYNY3axzaWg6Q7rkzvkp4q0vOcpM1kDO1qkS3san0MtMWM3Cy3V6vWmeASL7Uj9r_XcS3gkn-lz8rNTmb5sAV2e2huhjnJrndLx2mbSoBb8PU6ImW35gEHTwNm1qAGH3ZUMAKtx9K_7O-LR3OmuPu?purpose=fullsize)
