NMPC-Augmented Visual Navigation and Safe Learning Control for Large-Scale Mobile Robots

TL;DR

The paper addresses autonomous navigation and safe operation of a large-scale mobile robot on slip-prone terrain by integrating four modules: stereo vision-based pose estimation, real-time high-rate NMPC for wheel-motion commands, a learnable low-level actuation model (RSDNN) with adaptive safety, and a logarithmic safety supervisor. The approach jointly achieves real-time pose tracking, slip-aware high-level control, and robust actuation with stability guarantees, demonstrated by a uniform exponential stability result for the actuation subsystem and safe overall operation via the log-barrier. Experimental validation on a 6,000-kg LSMR shows the framework maintaining safety bounds while outperforming model-based and model-free baselines in actuation tracking, under both asphalt and soft-soil conditions. This work advances practical autonomy for heavy off-road robots by enabling interpretable learning-based control within rigorous stability and safety guarantees, suitable for high-stakes terrain scenarios.

Abstract

A large-scale mobile robot (LSMR) is a high-order multibody system that often operates on loose, unconsolidated terrain, which reduces traction. This paper presents a comprehensive navigation and control framework for an LSMR that ensures stability and safety-defined performance, delivering robust operation on slip-prone terrain by jointly leveraging high-performance techniques. The proposed architecture comprises four main modules: (1) a visual pose-estimation module that fuses onboard sensors and stereo cameras to provide an accurate, low-latency robot pose, (2) a high-level nonlinear model predictive control that updates the wheel motion commands to correct robot drift from the robot reference pose on slip-prone terrain, (3) a low-level deep neural network control policy that approximates the complex behavior of the wheel-driven actuation mechanism in LSMRs, augmented with robust adaptive control to handle out-of-distribution disturbances, ensuring that the wheels accurately track the updated commands issued by high-level control module, and (4) a logarithmic safety module to monitor the entire robot stack and guarantees safe operation. The proposed low-level control framework guarantees uniform exponential stability of the actuation subsystem, while the safety module ensures the whole system-level safety during operation. Comparative experiments on a 6,000 kg LSMR actuated by two complex electro-hydrostatic drives, while synchronizing modules operating at different frequencies.

Figures (6)

• Figure 1: Electro-hydrostatic drive, as a black-box model.
• Figure 2: The proposed control framework for the studied LSMR.
• Figure 3: Real-world validation on asphalt and soft soil.
• Figure 4: SLAM data for three scenarios on the soft soil terrain.
• Figure 5: Results of the logarithmic module (E(t) in meters).