This content will become publicly available on June 21, 2027

Title: Learning to Predict Mobile Robot Stability in Off-Road Environments
Navigating in off-road environments for wheeled mobile robots is challenging due to dynamic and rugged terrain. Traditional physics-based stability metrics, such as Static Stability Margin (SSM) or Zero Moment Point (ZMP) require knowledge of contact forces, terrain geometry, and the robot's precise center-of-mass that are difficult to measure accurately in real-world field conditions. In this work, we propose a learning-based approach to estimate robot platform stability directly from proprioceptive data using a lightweight neural network, IMUnet. Our method enables data-driven inference of robot stability without requiring an explicit terrain model or force sensing. We also develop a novel vision-based ArUco tracking method to compute a scalar score to quantify robot platform stability called C3 score. The score captures image-space perturbations over time as a proxy for physical instability and is used as a training signal for the neural network based model. As a pilot study, we evaluate our approach on data collected across multiple terrain types and speeds and demonstrate generalization to previously unseen conditions. These initial results highlight the potential of using IMU and robot velocity as inputs to estimate platform stability. The proposed method finds application in gating robot tasks such as precision actuation and sensing, especially for mobile manipulation tasks in agricultural and space applications. Our learning method also provides a supervision mechanism for perception based traversability estimation and planning.  more » « less
Award ID(s):
2150394
PAR ID:
10693772
Author(s) / Creator(s):
; ; ;
Publisher / Repository:
RSS 2025 Workshop on Resilient Off-road Autonomous Robotics
Date Published:
Subject(s) / Keyword(s):
Sponsoring Org:
National Science Foundation
More Like this
  1. null (Ed.)
    Abstract Robot-assisted healthcare could help alleviate the shortage of nursing staff in hospitals and is a potential solution to assist with safe patient handling and mobility. In an attempt to off-load some of the physically-demanding tasks and automate mundane duties of overburdened nurses, we have developed the Adaptive Robotic Nursing Assistant (ARNA), which is a custom-built omnidirectional mobile platform with a 6-DoF robotic manipulator and a force sensitive walking handlebar. In this paper, we present a robot-specific neuroadaptive controller (NAC) for ARNA’s mobile base that employs online learning to estimate the robot’s unknown dynamic model and nonlinearities. This control scheme relies on an inner-loop torque controller and features convergence with Lyapunov stability guarantees. The NAC forces the robot to emulate a mechanical system with prescribed admittance characteristics during patient walking exercises and bed moving tasks. The proposed admittance controller is implemented on a model of the robot in a Gazebo-ROS simulation environment, and its effectiveness is investigated in terms of online learning of robot dynamics as well as sensitivity to payload variations. 
    more » « less
  2. When a mobile robot is deployed in a field environment, e.g., during a disaster response application, the capability of adapting its navigational behaviors to unstructured terrains is essential for effective and safe robot navigation. In this paper, we introduce a novel joint terrain representation and apprenticeship learning approach to implement robot adaptation to unstructured terrains. Different from conventional learning-based adaptation techniques, our approach provides a unified problem formulation that integrates representation and apprenticeship learning under a unified regularized optimization framework, instead of treating them as separate and independent procedures. Our approach also has the capability to automatically identify discriminative feature modalities, which can improve the robustness of robot adaptation. In addition, we implement a new optimization algorithm to solve the formulated problem, which provides a theoretical guarantee to converge to the global optimal solution. In the experiments, we extensively evaluate the proposed approach in real-world scenarios, in which a mobile robot navigates on familiar and unfamiliar unstructured terrains. Experimental results have shown that the proposed approach is able to transfer human expertise to robots with small errors, achieve superior performance compared with previous and baseline methods, and provide intuitive insights on the importance of terrain feature modalities. 
    more » « less
  3. Tiny palm-sized aerial robots possess exceptional agility and cost-effectiveness in navigating confined and cluttered environments. However, their limited payload capacity directly constrains the sensing suite on-board the robot, thereby limiting critical navigational tasks in GPS-denied wild scenes. Common methods for obstacle avoidance use RGB cameras and LiDAR, which become ineffective in visually degraded conditions such as low visibility, dust, fog or complete darkness. Other sensors, such as RADAR, have high power consumption, making them unsuitable for tiny aerial robots. Inspired by bats, we propose Saranga, a low-power ultrasound-based perception stack that localizes obstacles using a dual sonar array. We present two key ideas to combat the low signal to noise ratio: physical noise reduction and a deep learning based denoising method. In the first idea, we find an optimal and practical way to block propeller induced ultrasound noise on the weak echoes. In the second idea, we generate and train a denoising neural network to utilize long-horizon for finding signal patterns under extreme amounts of uncorrelated noise. We generalize to the real-world with no real data for training. For the first time ever, we enable a palm-sized aerial robot to navigate in visually degraded conditions of smoke, darkness, and snow in a cluttered environment with thin and transparent obstacles using only on-board sensing and computation. We provide extensive real-world results to demonstrate the efficacy of our approach. 
    more » « less
  4. Tiny palm-sized aerial robots have exceptional agility and cost-effectiveness in navigating confined and cluttered environments. However, their limited payload capacity directly constrains the sensing suite onboard the robot, thereby limiting critical navigational tasks in Global Positioning System (GPS)–denied wild scenes. Common methods for obstacle avoidance use cameras and light detection and ranging (LIDAR), which become ineffective under visually degraded conditions such as low visibility, dust, fog, or darkness. Other sensors, such as radio detection and ranging (RADAR), have high power consumption, making them unsuitable for tiny aerial robots. Inspired by bats, we propose Saranga, a low-power, ultrasound-based perception stack that localizes obstacles using a dual sonar array. We present two key solutions to combat the low peak signal-to-noise ratio of −4.9 decibels: physical noise reduction and a deep learning–based denoising method. First, we present a practical way to block propeller-induced ultrasound noise on the weak echoes. The second solution is to train a neural network to use the long horizon of ultrasound echoes for finding signal patterns under high amounts of uncorrelated noise where classical methods were insufficient. We generalized to the real world by using a synthetic data generation pipeline augmented with limited real noise data for training. We enabled a palm-sized aerial robot to navigate under visually degraded conditions of dense fog, darkness, and snow in a cluttered environment with thin and transparent obstacles using only onboard sensing and computation. We provide extensive real-world results to demonstrate the efficacy of our approach. 
    more » « less
  5. Multi-robot cooperative control has been extensively studied using model-based distributed control methods. However, such control methods rely on sensing and perception modules in a sequential pipeline design, and the separation of perception and controls may cause processing latencies and compounding errors that affect control performance. End-to-end learning overcomes this limitation by implementing direct learning from onboard sensing data, with control commands output to the robots. Challenges exist in end-to-end learning for multi-robot cooperative control, and previous results are not scalable. We propose in this article a novel decentralized cooperative control method for multi-robot formations using deep neural networks, in which inter-robot communication is modeled by a graph neural network (GNN). Our method takes LiDAR sensor data as input, and the control policy is learned from demonstrations that are provided by an expert controller for decentralized formation control. Although it is trained with a fixed number of robots, the learned control policy is scalable. Evaluation in a robot simulator demonstrates the triangular formation behavior of multi-robot teams of different sizes under the learned control policy. 
    more » « less