Bio-inspired source seeking and obstacle avoidance on a palm-sized drone

From Shadows to Light: A Swarm Robotics Approach With Onboard Control for Seeking Dynamic Sources in Constrained Environments

“…The learning pipeline incorporated a training-insimulation phase prior to actual deployment. The authors in [11] introduced a bio-inspired solution, mimicking bacterial chemotaxis, to a source localization problem, but without employing any learning methodology. They deployed a single nano-UAV equipped with a temperature sensor to identify a heat source.…”

“…We use fully onboard controlled Crazyflie nano-drones in an indoor environment (unlike control from a central computer as in [11] or only on simulation as in [14] and [15]), a physical custom-made on-board sensor to implement scalar source sensing (differently from [13] where the measurements are calculated by a stored function), and another physical onboard sensor for sensing and avoiding an obstacle (which is not achieved in [13], [14] and [15]). Each drone senses the gradient via its sensor and retrieves a scalar value (the intensity of the source at its location), but it does not communicate this measurement to other drones (unlike the cases where measurements are communicated to peers for the estimation in [12], [13], [14] and [15]).…”

A Hybrid Simulation and Reinforcement Learning Algorithm for Enhancing Efficiency in Warehouse Operations

“…Simulation results confirmed the superiority of the proposed scheme in various layouts, highlighting the importance of cooperation among AGVs. Another approach was proposed by Elkunchwar et al [43], which addressed the autonomous source seeking capability for small unmanned aerial vehicles (UAVs) in challenging environments. Inspired by bacterial chemotaxis, a simple gradient-following algorithm was employed for source seeking while avoiding obstacles.…”

“…Ren and Huang [35] Balachandran et al [36] Gazebo Arslan and Ekren [37] Lewis et al [38] NVIDIA Isaac Sim Ho et al [39] Zhou et al [40] Cestero et al [41] Choi et al [42] Elkunchwar et al [43] Wang et al [44] Y. Ekren and Arslan [45] Arena…”

Tiny-PULP-Dronets: Squeezing Neural Networks for Faster and Lighter Inference on Multi-Tasking Autonomous Nano-Drones

“…Lightweight approaches better suited to nano-drones employ different sensors such as Time-of-Flight (ToF) ranging sensors [16], [17], [18]. These approaches can be implemented onboard and provide robust obstacle avoidance, even in unknown environments, with raw sensor readings or minimal onboard processing (e.g., 30 OP/s in [18]), such as simple buginspired lightweight state machines [16], [17]. However, in our competition, only a low-resolution monochrome monocular camera was allowed, narrowing the teams' effort only to visual-based approaches.…”