Vision-based Goal-Reaching Control for Mobile Robots Using a Hierarchical Learning Framework

TL;DR

The paper tackles safe goal-reaching for large-scale mobile robots operating on uncertain terrain by presenting a hierarchical, modular framework that combines vision-based perception, a smooth RL motion planner, a SCG-trained DNN inverse-dynamics model for in-wheel actuators, a model-based robust adaptive controller, and a mathematical safety supervisor. The approach integrates ORB-SLAM3 visual SLAM with a Q-learning/SARSA-based planner that outputs smooth velocity accelerations, while the DNN captures actuator dynamics and feeds a robust adaptive controller to ensure accurate trackability under slip. A Lyapunov-based stability analysis supports uniform exponential stability, and experiments on a 6,000 kg robot show centimeter-level goal-reaching accuracy (RMSE ≈ 3 cm) across asphalt and soft-soil surfaces, plus safe fault re-entry to a designated area. This work demonstrates a practical pathway to safe, autonomous operation of large, complex robots in harsh, partially known environments by cascading perception, planning, learning, and safety modules with formal guarantees.

Abstract

Reinforcement learning (RL) is effective in many robotic applications, but it requires extensive exploration of the state-action space, during which behaviors can be unsafe. This significantly limits its applicability to large robots with complex actuators operating on unstable terrain. Hence, to design a safe goal-reaching control framework for large-scale robots, this paper decomposes the whole system into a set of tightly coupled functional modules. 1) A real-time visual pose estimation approach is employed to provide accurate robot states to 2) an RL motion planner for goal-reaching tasks that explicitly respects robot specifications. The RL module generates real-time smooth motion commands for the actuator system, independent of its underlying dynamic complexity. 3) In the actuation mechanism, a supervised deep learning model is trained to capture the complex dynamics of the robot and provide this model to 4) a model-based robust adaptive controller that guarantees the wheels track the RL motion commands even on slip-prone terrain. 5) Finally, to reduce human intervention, a mathematical safety supervisor monitors the robot, stops it on unsafe faults, and autonomously guides it back to a safe inspection area. The proposed framework guarantees uniform exponential stability of the actuation system and safety of the whole operation. Experiments on a 6,000 kg robot in different scenarios confirm the effectiveness of the proposed framework.

Figures (6)

• Figure 1: Architecture of the RL motion planning.
• Figure 2: DNN training an inverse dynamics model of an actuator system used in many large-scale robots.
• Figure 3: The proposed framework in real time.
• Figure 4: Experimental surfaces: asphalt and soft-soil terrains.
• Figure 5: Goal reaching performance (meter).