Abstract: Mobile robots, especially those driving outdoors and in unstructured terrain, sometimes suffer from failures and errors in locomotion, like unevenly pressurized or flat tires, loose axes or de-tracked tracks. Those are errors that go unnoticed by the odometry of the robot. Other factors that influence the locomotion performance of the robot, like the weight and distribution of the payload, the terrain over which the robot is driving or the battery charge could not be compensated for by the PID speed or position controller of the robot, because of the physical limits of the system. Traditional planning systems are oblivious to those problems and may thus plan unfeasible trajectories. Also, the path following modules oblivious to those problems will generate sub-optimal motion patterns, if they can get to the goal at all. In this paper, we present an adaptive path planning algorithm that is tolerant to such degraded locomotion conditions. We do this by constantly observing the executed motions of the robot via simultaneously localization and mapping (SLAM). From the executed path and the given motion commands, we constantly on the fly collect and cluster motion primitives (MP), which are in turn used for planning. Therefore the robot can automatically detect and adapt to different locomotion conditions and reflect those in the planned paths.
Introduction
In ideal laboratory settings, a robot's wheel odometry is a reliable source of truth. However, in search and rescue or unstructured outdoor environments, hardware degradation is common. Tires lose pressure, axes become loose, or heavy payloads shift the center of gravity. Traditional navigation stacks, such as the standard ROS move_base, are often "oblivious" to these changes, issuing commands based on a theoretical kinematic model. If a robot has a flat left tire, a command to "go straight" might result in a veer to the left. Because wheel encoders might still show the wheels turning at the correct speed, the robot may not realize it is off-course until it collides with an obstacle. This research proposes an adaptive planning framework that makes the path planner aware of the robot's actual, current physical capabilities.
Methodology
The proposed system utilizes a three-step adaptive loop to ensure navigation success under degraded conditions:
1. Motion Primitive (MP) Generation
Instead of relying on a mathematical kinematic model, the system records actual robot motions. By pairing control inputs (velocity commands) with the resulting path measured via SLAM (rather than wheel odometry), the robot builds a library of its actual capabilities. Each primitive is defined by its geometric properties, such as arc length and curvature.
2. Clustering and Planning
To maintain computational efficiency, the system uses Affinity Propagation to cluster hundreds of recorded motions into a representative set of "Motion Primitives." These primitives serve as the edges for an A* search algorithm. This ensures that any planned path is, by definition, something the robot has recently proven it can execute.
3. Adaptive Feedback Loop
As the robot executes a plan, it constantly evaluates its performance. If the robot fails to reach the expected pose of a specific MP, that primitive is "downvoted" via a voting system. If locomotion conditions change mid-mission (e.g., a tire failure), the system detects the inconsistency, downvotes obsolete MPs, and incorporates new ones that reflect the current state. A heuristic punishment is also applied to cells where the robot has previously failed, encouraging the planner to explore paths compatible with current physical limitations.
Experimental Results
The researchers evaluated the approach using a Clearpath Jackal mobile robot equipped with a Velodyne VLP16 LIDAR and an IMU. Localization was provided by the BLAM (Berkeley Localization and Mapping) package to ensure a ground truth independent of wheel odometry. The system was tested against the standard ROS move_base (using the Dynamic Window Approach, DWA) across four scenarios: Normal, Right Constrained (simulated failure), Left Constrained, and Overload.
- Success Rate: In all degraded conditions, the standard
move_basefailed to reach the goal. The adaptive method succeeded in every test. - Accuracy: In the "Right Constrained" test, the standard navigator ended 3.54 meters away from the goal, while the proposed method reached within 0.73 meters.
- Versatility: The system proved effective for both "veering" (unbalanced degradation) and "sluggishness" (balanced degradation caused by overloading).
Conclusion
By using SLAM as a truth source to validate odometry, the proposed framework allows robots to quantify and adapt to locomotion degradation without manual physical modeling. Future work will focus on moving toward a 100% online clustering system to enable true autonomy in unknown environments without the need for a pre-obtained primitive library.
Summary updated