Gayan Kahandawa

Prior to identifying optimal restraint or containment systems to prevent people using manual wheelchairs sliding or tipping on large public transit buses, the forces that destabilise them must be understood. This pilot study documented the (1) forces that wheelchairs experience on buses, (2) movement of wheelchairs on buses, and (3) differences between normal operating forces and the forces that lead to a wheelchair sliding or tipping. Methods: Data on the forces that acted on three manual wheelchairs were gathered using accelerometers across six bus routes with varied terrain. Data ranges were summarised using MATLAB, and Rainflow analysis generated graphs for visual inspection. The points at which wheelchairs began to slide, and tip were modelled and simulated in 3D in SOLIDWORKS and MSC Adams software respectively. Direct comparisons were then made between normal operating forces and the forces that lead to slide or tip. Real-world driving forces that act on wheelchairs under typical driving conditions ranged from -0.62 g to 0.44 g. The points at which wheelchairs began to slide or tip were documented. A table showing normal operating forces and the forces that lead to slide or tip was prepared confirming slide and tip can occur across regular manoeuvres, although such forces are rarely generated. Conclusion: These pilot data inform bus designers, operators, and transport policymakers of the forces that act on manual wheelchairs that produce slide or tip so that driver behaviour can be monitored and restraint or containment systems can be continually improved, thereby promoting safety for all passengers.

Reliable navigation in unstructured outdoor environments requires autonomous ground robots to reason about both the geometric structure and the semantic properties of the surrounding terrain. Existing mapless navigation methods, including the Agoraphilic-3D (A3D) algorithm, effectively account for terrain geometry and robot stability but lack awareness of terrain type, limiting their applicability in environments where surface characteristics such as traction, material, or safety constraints significantly influence mobility. This work introduces Agoraphilic 3D-Terrain Aware (A3D-TA), a novel extension of A3D that integrates real-time terrain classification with 3D geometric free-space analysis to enable semantically informed navigation decisions. The proposed framework employs a segmentation model to identify terrain classes and fuses these labels with depth measurements to generate semantically annotated point clouds. Sector-wise traversability is then computed by combining terrain geometry, terrain-type priority weighting, and near-field safety considerations. A3D-TA further incorporates a terrain-aware velocity adaptation mechanism that adjusts driving speed according to surface characteristics. The experimental results validate the effectiveness of the proposed A3D-TA framework in real-world outdoor navigation when compared against baseline methods.

This paper presents Agoraphilic-3D Dynamic Obstacle-aware (A3D-DO), a unified dynamic-obstacle-aware navigation framework for autonomous ground robots operating on uneven terrain, grounded in the free-space-driven Agoraphilic navigation principle. The proposed framework extends free-space-based motion selection beyond static geometric reasoning by explicitly modelling how dynamic obstacles will constrict or interrupt future navigable space. Central to A3D-DO is a predictive formulation that estimates the time-varying evolution of sector-wise free space under both collinear and crossing obstacle motion patterns. This prediction is tightly coupled with slope-based stability reasoning, allowing the robot to evaluate navigable directions that are not only geometrically feasible but also stable and temporally safe. By fusing static terrain traversability with predicted dynamic free-space availability, A3D-DO produces a unified static-dynamic free-space representation that enables anticipatory and stability-consistent navigation decisions. The effectiveness of the proposed framework is demonstrated through extensive simulation studies and real-world experiments on uneven terrain with diverse dynamic-obstacle behaviours, where A3D-DO consistently achieves higher success rates and safer navigation compared to conventional dynamic-obstacle avoidance methods.

Autonomous ground robot navigation in unstructured, mapless, and uneven terrains remains a challenging task due to variable slopes, environmental occlusions, and the need to preserve robot stability. To address these challenges, this paper presents Agoraphilic-3D (A3D), a novel path planning algorithm designed for real-world, uneven terrain navigation with explicit consideration of stability constraints. A3D extends the free space based navigation concept of the original Agoraphilic algorithm, advancing it from 2D environment navigation to 3D environment navigation. The proposed algorithm introduces a Square Sector Map (SS-Map), a robot-centric grid representation of 3D point cloud data, which enables efficient terrain analysis and free space determination. An unsupervised clustering-based occlusion filtering method is incorporated to remove ambiguous regions, producing cleaner data for terrain feature extraction. Using the SS-Map together with the filtered data, the terrain slope is assessed relative to the robot’s stability constraints, defined by its maximum allowable pitch and roll angles, and encoded into a 1D free space histogram. Based on this representation, free-space attractive forces are derived to guide navigation through the most stable regions, while a steering factor mechanism simultaneously biases the trajectory toward the goal. Extensive experiments in both simulation and real-world outdoor environments demonstrate the effectiveness of the A3D algorithm. The results highlight its robustness in identifying terrain variations, filtering unstable regions, and generating safe, traversable paths without reliance on prior maps. Overall, A3D provides a practical and adaptable framework for autonomous ground navigation in challenging domains such as exploration, agriculture, and search and rescue.