ABSTRACT. This paper proposes a multi-camera visual predictive control strategy for a mobile manipulator allowing to position the end-effector camera with respect to a landmark. Several issues are considered: (i) the visual landmark possible loss during navigation, (ii) the realization of large displacements which implies a large prediction horizon and impacts the closed-loop stability, (iii) the robot's high redundancy which may lead to a large search space and potential non-relevant solutions, (iv) the processing time. To cope with these challenges, the proposed strategy relies on (i) the use of two complementary cameras, (ii) the definition of a cost function depending on both the vision-based task and the manipulability, (iii) the integration of constraints allowing to prioritize the former against the latter. The strategy has been simulated using ROS and Gazebo and compared with our previous work, showing its efficiency.

ABSTRACT. This paper presents a novel formulation of motion planning for multi-legged walking robots. In the proposed method, a single-step motion is formulated as a nonlinear equation problem (NLE): including kinematic, stability, and collision constraints. For the given start and goal configurations, the robot's path is parametrized as a Bézier curve in the configuration space. The resulting NLE is solved using Levenberg-Marquardt optimization implemented using a sparse matrix solver. We propose handling the trigonometric kinematic constraints with the polynomial path parametrization. A relaxation of the constraint is used while guaranteeing a desired tolerance along the planned path. Although the proposed method does not explicitly optimize any criterion, it produces high-quality paths. The method is deployed in transforming a sequence of discrete configurations produced by a step sequence planner into a valid path for a multi-legged walking robot in challenging planning scenarios where a regular locomotion gait cannot be used because of sparse footholds.

ABSTRACT. Safe mapless navigation of mobile robots in unknown and human populated areas is integral for increasing their usage in our daily lives. In this paper, we consider how such a behavior can be exhibited by a mobile robot and introduce Social APF-RL (Artificial Potential Functions with Reinforcement Learning). Social APF-RL extends our previously presented approach APF-RL in which the strengths of artificial potential functions (APF) with deep reinforcement learning are combined so that the robot learns how to adjust the input parameters of the APF controller. With Social APF-RL, the model is extended to accomodate the presence of humans and to respect their comfort zones while navigating. Our experimental results including both simulation and real-life scenarios demonstrate that differing from the classical navigation methods or social navigation methods, the robot can navigate successfully on its own even in complex scenarios with moving entities while maintaining social distance to humans encountered. Hence, it has better applicability in real-life scenarios. For future work, we plan to use the proposed approach in human following while adhering to social distance norms.

ABSTRACT. This work focuses on the automatic control of the length of a tether that links an underwater vehicle to the surface, with the objective to prevent the tether from becoming taut or getting entangled due to too much length being deployed. The solution proposed here consists of equipping the tether with a balanced buoy-ballast system that gives the cable a V- shape in the vicinity of the vehicle. This system offers a passive compliance by smoothing the movements of the tether and damping external disturbances. The tether length is adjusted by an active feeder on the surface, whose control relies on the reading of a flex sensor embedded in the V-shape portion of the cable. The experiments conducted on a real ROV in a pool allowed validating this mechatronic compliant-actuated system, which can adapt to the movements of the underwater vehicle while it executes longitudinal and curved trajectories.

ABSTRACT. Multi-robot formations have numerous applications, such as cooperative object transportation in smart warehouses. Here, robots must deliver objects in formation while avoiding intra- and inter-formation collisions. This requires solutions to multi robot task assignment, formation generation, rigid formation maintenance, and route planning. In this paper, we present a cooperative multi-formation object transportation system which explicitly handles inter-formation collisions. For formation generation, we propose a distributed motion planning approach which combines artificial potential field methods and leader-follower based control. For formation planning, we present a heuristic search-based algorithm which uses convex segmentation techniques, and extend the minimum snap method to synthesise smooth trajectories while maintaining the formation. We also propose a variant of the dynamic window approach to avoid collisions between formations. We demonstrate the efficacy of our approach in simulation.

ABSTRACT. Unmanned Aerial Vehicles (UAVs) are highly manoeuvrable platforms that are becoming popular to map and localise the source of chemical substances in the environment. However, their propellers disturb the environment and may seriously impact the measurements and the expected results in such applications. Some works have already addressed this problem, but only for specific cases and proposing different solutions about the best way to mitigate this problem. This work studies the impact of propellers for multiple sensor positions and three UAVs of different sizes, in an Odour Source Localisation (OSL) problem under field conditions and compares the results with a non-disturbing setup. The experiments show a significant impact on the measured signals from asymmetrical frames and from the motion of the agent, with the best results achieved from sensors acquiring data on intake air regions.

ABSTRACT. In this paper, we propose a novel vector-field based guidance law for an unmanned aircraft (UA) to track a stationary and a moving target while maintaining a desired standoff distance under constant vehicle speed conditions in the absence and presence of wind. The guidance law achieves the convergence of the path of UA to the desired standoff distance as well as its heading angle to the desired heading angle in finite-time. We analyze the theoretical properties of the proposed guidance law using Lyapunov theory and evaluate the performance of the proposed guidance law through simulations and hardware experiments.

ABSTRACT. This paper addresses the problem of detecting humans in RGB and Thermal (long-wave IR) images taken by cameras mounted onboard a mobile robot. Human/Pedestrian detection is currently one of the most pertinent object detection problems, mainly due to safety concerns in autonomous vehicles. The majority of approaches apply deep-learning techniques based solely on RGB images. However, they have a few shortcomings, namely that during foggy weather, nighttime, and low-light scenarios, these images may not contain sufficient information. To address these issues, this work studies the use of thermal cameras as a complementary source of information for human detection in indoor and outdoor environments. The proposed approach uses YOLOv5 to detect pedestrians in both thermal and RGB images. Moreover, the different modalities are combined using early and late fusion techniques. Evaluation of the proposed approach is carried out in the FLIR Aligned dataset and in a new in-house dataset. Results indicate that the use of fusion techniques highlights a promising way to improve the overall performance in this application domain.