Control of Rough Terrain Vehicles Using Deep Reinforcement Learning

  title={Control of Rough Terrain Vehicles Using Deep Reinforcement Learning},
  author={Viktor Wiberg and Erik Wallin and Tomas Nordfjell and Martin Servin},
  journal={IEEE Robotics and Automation Letters},
We explore the potential to control terrain vehicles using deep reinforcement in scenarios where human operators and traditional control methods are inadequate. This letter presents a controller that perceives, plans, and successfully controls a 16-tonne forestry vehicle with two frame articulation joints, six wheels, and their actively articulated suspensions to traverse rough terrain. The carefully shaped reward signal promotes safe, environmental, and efficient driving, which leads to the… 
Risk-Aware Off-Road Navigation via a Learned Speed Distribution Map
This work proposes a new representation of traversability based exclusively on robot speed that can be learned from data, offers interpretability and intuitive tuning, and can be easily integrated with a variety of planning paradigms in the form of a costmap.
Analytic Solutions for Wheeled Mobile Manipulator Supporting Forces
When a mobile manipulator’s wheel loses contact with the ground, the manipulator may overturn, causing material damage, and in the worst case, putting human lives in danger. The overturning stability


Deep Reinforcement Learning for Safe Local Planning of a Ground Vehicle in Unknown Rough Terrain
A deep reinforcement learning approach for local planning in unknown rough terrain with zero-range to local-range sensing, achieving superior results compared to potential fields or local motion planning search spaces methods.
Robot Navigation of Environments with Unknown Rough Terrain Using deep Reinforcement Learning
This paper has developed and trained a DRL network that uses raw sensory data from the robot's onboard sensors to determine a series of local navigation actions for a mobile robot to execute and successfully tested in several unique 3D simulated environments.
Learning quadrupedal locomotion over challenging terrain
The presented work indicates that robust locomotion in natural environments can be achieved by training in simple domains.
Control of Robotic Vehicles with Actively Articulated Suspensions in Rough Terrain
A method to control actively articulated suspensions to enhance rover tipover stability using a quasi-static model, and optimized on-line, relies on estimation of wheel-terrain contact angles.
Learning-Based Methods of Perception and Navigation for Ground Vehicles in Unstructured Environments: A Review
A review on the recent contributions in the robotics literature adopting learning-based methods to solve the problem of environment perception and interpretation with the final aim of the autonomous context-aware navigation of ground vehicles in unstructured environments is presented.
High-Dimensional Continuous Control Using Generalized Advantage Estimation
This work addresses the large number of samples typically required and the difficulty of obtaining stable and steady improvement despite the nonstationarity of the incoming data by using value functions to substantially reduce the variance of policy gradient estimates at the cost of some bias.
Design, analysis and experiment of an eight-wheel robotic vehicle with four-swing arms
The experiment results demonstrate that the robotic vehicle can surmount an obstacle 2.29 times the height of the wheel radius, which verifies the feasibility of this new configuration of an all-terrain robotic vehicle with four-swing arms.
DeepMimic: Example-Guided Deep Reinforcement Learning of Physics-Based Character Skills
This work shows that well-known reinforcement learning methods can be adapted to learn robust control policies capable of imitating a broad range of example motion clips, while also learning complex recoveries, adapting to changes in morphology, and accomplishing user-specified goals.
Force Control for Active Chassis Balancing
This paper presents the realization of the worldwide first automated walking excavator chassis. To this end, the authors build a new generation of high-performance hydraulic valves with integrated
ALLSTEPS: Curriculum‐driven Learning of Stepping Stone Skills
The importance of a curriculum for efficient learning is demonstrated and four possible curriculum choices compared to a non‐curriculum baseline are evaluated and shown to be viable.