AMR Dynamic Obstacle Avoidance Patents 2026 — PatSnap Eureka
AMR Dynamic Obstacle Avoidance: 2026 Technology Landscape
Classical geometric planners are being augmented by deep reinforcement learning pipelines, enabling AMRs to navigate genuinely unstructured, human-populated environments. This report maps 62 retrieved records spanning 2004–2026.
Four Technical Families Shaping AMR Obstacle Avoidance
Dynamic obstacle avoidance (DOA) for AMRs requires robots to respond in real time to obstacles whose positions, velocities, and trajectories change over time. Within this dataset, four broad technical families are visible: velocity-space and geometric planning methods, sampling-based and graph-search planners, reinforcement and deep learning methods, and sensor fusion and perception pipelines.
Velocity-space methods model obstacle motion as velocity obstacles (VO), generalized velocity obstacles (GVO), dynamic window approaches (DWA), or artificial potential fields (APF), translating obstacle kinematics directly into robot control commands. Sampling-based methods such as RRT, RRT*, A*, and D* Lite construct or continuously re-plan collision-free paths as new obstacles are detected.
Deep reinforcement learning methods — including DQN, PPO, SAC, and LSTM-based controllers — learn avoidance policies through environmental interaction. The challenge motivating virtually all retrieved records is the same: static-environment planning assumptions break down when people, vehicles, or other robots share the workspace, requiring both accurate obstacle state estimation and planners capable of reasoning about future positions.
Publication dates in this dataset span from 2004 to 2026, revealing three phases: a foundational phase (2004–2014) establishing core algorithmic primitives, a development phase (2015–2020) driven by ROS and LiDAR commoditization, and a convergence phase (2021–2026) marked by hybrid architectures and increased commercial patent filings. Among 11 patent records with assignee metadata in this dataset, India accounts for 6 filings, followed by the US with 3.
Patent Filing Activity and Algorithm Cluster Distribution
Patent filing activity in this dataset accelerated after 2020, with the majority of commercial filings appearing in 2024–2026. Algorithm distribution across retrieved records shows deep reinforcement learning and sensor fusion methods gaining ground over classical geometric planners.
Algorithm Cluster Distribution Across Retrieved Records (Dataset Snapshot)
Sensor fusion and DRL-based methods together represent the majority of records in this dataset from 2020 onward, while velocity-space methods anchor foundational filings from the 2004–2018 period.
↗ Click bars to explorePatent Filing Activity by Phase — AMR Obstacle Avoidance (Dataset Snapshot)
Commercial patent filings in this dataset are concentrated in 2021–2026, with 8 of 11 assignee-identified patents filed after 2020, reflecting accelerating commercialization of AMR navigation technology.
↗ Click bars to exploreWhere AMR Dynamic Obstacle Avoidance Is Being Deployed
Retrieved records across this dataset reveal AMR dynamic obstacle avoidance being applied across four major domains: warehouse logistics, autonomous ground vehicles, unmanned aerial vehicles, and service robots in human-shared environments.
Warehouse Logistics & Industrial AMRs
Zebra Technologies Corporation’s 2020 US patent addresses mobile automation apparatus navigation using dynamic perimeter regions calibrated by localization confidence for warehouse environments. A 2023 literature record applies hybrid APF-MPC to AGV fleets at automated container terminals, handling both static lane boundaries and dynamic conflicting vehicles simultaneously.
Industrial AutomationAutonomous Ground Vehicles
A 2022 literature record applies DDPG with V2X sensor data to extend perception beyond local sensors for connected autonomous vehicles. A 2026 IN patent by Mandapalli Mahikumar uses Anytime Dynamic A* replanning with ML-based navigation efficiency evaluation for autonomous vehicle routing. Shanghai Electric-Hydrogen’s 2026 CN patent targets autonomous transport robots with multi-objective path scoring for safety, smoothness, and efficiency.
Autonomous VehiclesUnmanned Aerial Vehicles
Everseen Limited’s 2024 US patent uses depth maps re-scaled to global environment maps to trace routes avoiding both static and dynamic obstacles in aerial navigation systems. The 2022 DPMPC-Planner literature record separates static mapping from dynamic object representation using computer vision, then applies chance-constrained MPC for reactive UAV avoidance in complex environments.
UAV NavigationService Robots & Human Environments
A 2020 literature record on DRL-based semantic navigation addresses mobile robots operating among unpredictable pedestrians in service contexts. Honda Motor Co., Ltd.’s trajectory planner patents (US, 2021 and 2024) address all-terrain vehicles and wheelchairs operating in unknown dynamic environments. The RL-DOVS 2022 record represents environmental dynamism as a robocentric velocity space fed to an RL agent for human-shared navigation.
Human-Robot InteractionLeading Assignees in AMR Obstacle Avoidance — Dataset Snapshot
Among 11 patent records with assignee metadata in this dataset, Delta Electronics Int’l (Singapore) Pte Ltd and Indian Institute of Technology Madras each hold 2 filings in retrieved records, tied with Honda Motor Co., Ltd. at 2 filings, making them the most active individual assignees in this dataset.
Top Assignees by Filing Count in Retrieved Records (Dataset Snapshot)
↗ Click bars to exploreDelta Electronics Int’l (Singapore) Pte Ltd
Delta Electronics Int’l (Singapore) Pte Ltd holds 2 patent filings in retrieved records: one US filing and one SG filing, both from 2025, covering autonomous mobile robot operating methods. The patents claim explicit logic for distinguishing dynamic obstacle direction — same vs. opposite to the AMR — and adjusting avoidance strategy based on obstacle speed thresholds, including maintaining safe distance or re-routing. This direction-aware behavioral approach is described as a step beyond simple collision avoidance toward socially-aware navigation in this dataset.
Singapore / United StatesIndian Institute of Technology Madras
Indian Institute of Technology Madras holds 2 patent filings in retrieved records: one WO filing and one IN filing, both from 2025, covering a route planning method for autonomous mobile objects. The patents generate virtual triangulated 3D geometric casings around obstacles, enabling more accurate 3D free-space computation compared to 2D cost maps, applicable to both ground and aerial platforms. The WO filing signals intent to seek international protection for this 3D obstacle representation approach.
India — IN / International WOFive Converging Technical Directions in AMR Obstacle Avoidance (2024–2026)
The most recent filings and publications in this dataset (2024–2026) point to five converging technical directions: hybrid DRL and classical planner architectures, direction- and speed-aware obstacle classification, 3D geometric obstacle representation, multi-modal sensing with AI classification, and transport robot specialization.
Hybrid DRL + Classical Planner Architectures
Records from 2021–2022 established the theoretical bridge between DRL local planners and conventional global planners via intermediate waypoint generators. The 2025 IN patent on BiRRT-SARSA hybrid dynamic path finding and the All-in-One DRL control switch architecture (2022) illustrate this convergence moving into products. R&D investment should focus on the integration layer — waypoint generators, control switches, and planner arbitration logic — rather than either paradigm alone.
Direction- and Speed-Aware Obstacle Classification
Delta Electronics’ 2025 US and SG patents introduce fine-grained behavioral rules: if a dynamic obstacle moves in the same direction at lower speed, re-route; if opposite, determine pathway width and execute a passing strategy. This behavioral specificity goes beyond simple collision avoidance and is directly relevant to human-facing service robots. Freedom-to-operate assessments in this sub-domain are recommended for firms deploying AMRs in human-shared environments.
Classical Geometric Planners vs. Deep Reinforcement Learning Methods
Click any row to explore further.
| Dimension | Classical Geometric Planners (VO/DWA/RRT) | Deep Reinforcement Learning (DQN/PPO/SAC) |
|---|---|---|
| Core Mechanism | Model obstacle kinematics as velocity obstacles or forbidden velocity cones; select robot action outside forbidden set | Learn avoidance policy by rewarding collision-free goal-reaching behavior through environmental interaction |
| Obstacle Handling | Explicitly models obstacle velocity vectors; fails when static-environment planning assumptions are violated in dense dynamic scenes | Handles unpredictable pedestrians and multi-obstacle scenarios; struggles with long-range navigation consistency |
| Key Algorithms | VO, GVO, DWA, APF, A*, RRT, RRT*, D* Lite, PRM, hybrid A* | DQN, PPO, SAC, DDPG, LSTM-based controllers, Q-learning |
| Sensor Input | LiDAR, laser range finders, Kalman filter-tracked obstacle state estimates | Raw sensor observations, robocentric velocity space representations, semantic inputs |
| Compute Requirements | Lower; geometric computation suitable for edge hardware in ROS navigation stacks | Higher training overhead; inference cost depends on policy complexity and hardware |
| Representative Records | RRT*+GVO (2018 literature), OP-PRM+D* Lite (2022 literature), IIT Madras 3D casing patents (2025) | SAC+ROS (2021 literature), DRL-waypoint bridge (2021 literature), DQN/PPO/SAC survey of 34 studies (2023) |
| Commercial Filings (dataset) | Zebra Technologies 2020 US, Delta Electronics 2025 US/SG, Chennai Institute of Technology 2025 IN | Tata Consultancy Services 2026 IN (VO+DRL hybrid), Honda Motor 2021/2024 US trajectory planners |
| Hybrid Integration Trend | Used as global planner backbone; combined with DRL local planners via waypoint generator bridge layers | Used as local planner; depends on classical global planner for long-range route generation |
Frequently Asked Questions: AMR Dynamic Obstacle Avoidance Patents
Within this dataset, four broad technical families are visible: (1) velocity-space and geometric planning methods (VO, GVO, DWA, APF); (2) sampling-based and graph-search planners (RRT, RRT*, A*, D* Lite, PRM); (3) reinforcement and deep learning methods (DQN, PPO, SAC, DDPG, LSTM); and (4) sensor fusion and perception pipelines combining LiDAR, cameras, and Kalman filter tracking.
Among the 11 patent records with assignee and jurisdiction metadata in this dataset, India (IN) leads with 6 filings, followed by the United States (US) with 3 filings, international (WO) with 2 filings, and Singapore (SG) and China (CN) each with 1 filing. India’s count reflects a broad base of academic institutions, individual inventors, and technology services firms.
The newest filing in this dataset is a CN patent from Shanghai Electric-Hydrogen Intelligent Transport Robot Technology Co., Ltd. (also known as Shanghai Dianqing Zhiyun Robot Technology Co., Ltd.), published in March 2026, targeting environment perception and obstacle avoidance for autonomous transport robots with multi-objective path scoring.
Records from 2021–2022 established a bridge using intermediate waypoint generators that connect DRL local planners with conventional global planners, improving safety and path smoothness. By 2025–2026, patent filings such as the BiRRT-SARSA hybrid (Chennai Institute of Technology, 2025 IN) show this approach moving into products. The 2022 All-in-One DRL control switch architecture also illustrates this convergence.
Delta Electronics Int’l (Singapore) Pte Ltd filed both a US patent and a SG patent in 2025 covering autonomous mobile robot operating methods. These patents introduce fine-grained behavioral rules: if a dynamic obstacle moves in the same direction at lower speed, re-route; if opposite, determine pathway width and execute a passing strategy. This direction- and speed-aware obstacle classification goes beyond simple collision avoidance toward socially-aware navigation.
Multiple records identify dynamic obstacle velocity estimation as the critical input to effective avoidance: the 2022 ROS AMR literature paper, the Tata Consultancy Services 2026 IN patent, and the Delta Electronics 2025 US patent all address this. Techniques such as Extended Kalman Filter combined with vision tracking and laser range finders are used to estimate obstacle velocities and update cost maps. IP opportunities are noted for EKF/IMU fusion for velocity estimation at low computational cost on edge hardware.
Data and insights on this page are based on a limited patent and literature dataset and are for reference only. Figures may not represent the complete technology landscape.