On-Board Human-Aware Navigation for Indoor Resource-Constrained Robots: A Case-Study with the Ranger

Introducing simple robotic platforms into domes- tic environments is faced with the challenge of social accept- ability. Therefore human-aware navigation is a must for robots operating in environments shared with human users. In this work, we focus on the human-aware navigation problem in a structured environment for a robot with limited sensing and constrained maneuvering called Ranger. The Ranger is a simple domestic robotic platform designed for interacting with children. The system combines person detection and tracking —which is the result of fusing laser-scan and depth-image based detectors provided by an RGB-D camera—, basic autonomous navigation and the concept of personal space. We rely only on the on-board sensors for mapping, localization, human tracking, and navigation. Systematic experiments are carried out with a real robot in the presence of a human in order to compare our human-aware navigation with a non human-aware simple approach. The results show that human-aware navigation is able to achieve trajectories which are respecting the personal spaces of the human and are thus more acceptable for the users.

Bio-inspired and probabilistic algorithms for distributed odor source localization using mobile robots

Thomas Lochmatter

We compare six different algorithms for localizing odor sources with mobile robots. Three algorithms are bio-inspired and mimic the behavior of insects when exposed to airborne pheromones. Two algorithms are based on probability and information theory, and infer the source location by probabilistically merging concentration measurements at different positions in the environment. The last algorithm is a multi-robot algorithm based on a crosswind line formation. The algorithms are mainly compared with respect to their distance overhead – a metric directly related to the speed of an algorithm – and their success rate. The thesis also reports on the influence of various environmental and algorithmic parameters, and compares the algorithms' requirements regarding sensors, self-localization, maps, and other information. Systematic experiments under laminar flow conditions were carried out with real robots in an 18m long wind tunnel. The robots were thereby equipped with an ethanol sensor and a wind direction sensor, and could – if the algorithm required it – access their current position. Overall, more than 500 experimental runs with teams of up to 5 robots were carried out in this wind tunnel. Similar experiments were also carried out in simulation. Over 5000 runs were carried out in a realistically calibrated multi-robot simulator. Odor was thereby simulated as set of filaments that are transported by advection, an approach that generates the intermittence and stochasticity of real plumes. Additional, more than 10000 runs were carried out using body-less simulators under various plume models. Simulation runs were mostly used to quantify the influence of various parameters on the performance of the algorithms. Finally, the thesis also provides theoretical insights into the bio-inspired algorithms, and a general theoretical model for probabilistic odor source localization. For the latter, a number of potential real-world scenarios are discussed on the example of a simplified train station environment. None of the algorithms is strictly superior to all other algorithms. While the probabilistic algorithms offer more flexibility and a slightly better performance, the bio-inspired algorithms are much less CPU and memory intensive, and could therefore be deployed on extremely small and limited robotic platforms. Using multiple robots (with or without collaboration) for odor source localization was found to improve the performance under certain conditions only. The crosswind formation algorithm with 3 robots yielded excellent results, but the multi-robot experiments with the bio-inspired algorithms were hardly better than their single-robot counterparts. The thesis provides reasons for this, and discusses alternatives.

EPFL2010

Brain-coupled Interaction for Semi-autonomous Navigation of an Assistive Robot

Ricardo Andres Chavarriaga Lozano, Xavier Perrin, Roland Siegwart

This paper presents a novel semi-autonomous navigation strategy designed for low throughput interfaces. A mobile robot (e.g. intelligent wheelchair) proposes the most probable action, as analyzed from the environment, to a human user who can either accept or reject the proposition. In case of refusal, the robot will propose another action, until both entities agree on what needs to be done. In an unknown environment, the robotic system first extracts features so as to recognize places of interest where a human-robot interaction should take place (e.g. crossings). Based on the local topology, relevant actions are then proposed, the user providing answers by the mean of a button or a brain-computer interface (BCI). Our navigation strategy is successfully tested both in simulation and with a real robot, and a feasibility study for the use of a BCI confirms the potential of such interface.

Elsevier2010

The Effect of the Environment in the Synthesis of Robotic Controllers: A Case Study in Multi-Robot Obstacle Avoidance using Distributed Particle Swarm Optimization

Ezequiel Leonardo Di Mario, Alcherio Martinoli, Inaki Navarro Oiza

The ability to move in complex environments is a fundamental requirement for robots to be a part of our daily lives. While in simple environments it is usually straightforward for human designers to foresee the different conditions a robot will be exposed to, for more complex environments the human design of high-performing controllers becomes a challenging task, especially when the on-board resources of the robots are limited. In this article, we use a distributed implementation of Particle Swarm Optimization to design robotic controllers that are able to navigate around obstacles of different shape and size. We analyze how the behavior and performance of the controllers differ based on the environment where learning takes place, showing that different arenas lead to different avoidance behaviors. We also test the best controllers in environments not encountered during learning, both in simulation and with real robots, and show that no single learning environment is able to generate a behavior general and robust enough to succeed in all testing environments.

MIT Press2013

Bio-inspired and probabilistic algorithms for distributed odor source localization using mobile robots

Thomas Lochmatter

EPFL2010