List all unit tests and overall system test carried out for experimentation of the system. List any findings and design changes made from your analysis of test results and other data obtained from the experimentation.
Unit and System Testing
This week, testing focused on validating the navigation stack across multiple environments and durations. For mapping, the team ran Cartographer SLAM across several different room configurations to evaluate map quality and consistency. Output maps were reviewed for accuracy in both open spaces and constrained corridors, with attention to how well scan-matched features held up as the robot traversed room boundaries. Navigation endurance tests were also conducted at varying durations 1, 10, and 15 minute runs to observe how localization and path tracking degraded over time and whether any systematic drift or replanning failures emerged during longer operation.
Design Changes from Test Results
Analysis of the navigation test runs revealed that the previous controller, Regulated Pure Pursuit (FollowPath/RPP), exhibited goal convergence issues near the destination, where the robot would oscillate and fail to settle cleanly. In response, the local controller was migrated from RPP to the Dynamic Window Approach (DWB) planner within Nav2. DWB samples a set of candidate trajectories at each control cycle and scores them across multiple critics including path alignment, goal heading, and obstacle clearance, giving the controller more flexibility for clean stopping and heading recovery. Critic weights were tuned to make the controller more aggressive in correcting heading error, which reduced the oscillation behavior observed during testing.
Separately, odometry performance was identified as a contributing factor to map drift during longer runs. The SparkFun OTOS sensor exposes linear and angular scalar calibration parameters that were adjusted based on measured ground truth. The linear scalar was tuned by driving a known tape-measured distance and comparing against OTOS-reported displacement, and the angular scalar was tuned by commanding a full 360-degree rotation and verifying heading return to within tolerance. Tightening these scalars reduced odometry drift and improved Cartographer’s pose estimate stability during extended navigation runs.
Planned Future Testing
The next phase of system testing will focus on end-to-end delivery runs using a pre-saved map of the target environment. The plan is to define a waypoint sequence in Nav2 that represents a realistic delivery mission: starting from a home position, navigating to a pickup location, proceeding to a delivery destination, and returning back to the origin. This will exercise the full autonomy stack in a mission-relevant context rather than isolated navigation trials. Key metrics to evaluate during these runs include total mission completion time, localization consistency across the full route, and how reliably the robot reaches and confirms each waypoint before proceeding. Any failures in goal convergence, replanning, or behavioral FSM state transitions triggered by the KR260 person detections during the run will be logged and used to drive further tuning of the DWB critic weights and Nav2 parameters.