Sensing and perception are fundamental problems for mobile robotics in general and for field robotics in particular. The team has studied for long time the integration of uncommon sensing modalities in robotics, e.g. olfaction, achieving excellent results in the area. During the evaluation period, particular focus on cooperative search and exploration algorithms using this modality was provided, but the experimental validation was mainly restricted to indoor environments. Efficient olfactory sensing in large scale natural environments and all the related perception and decision issues are still an open problem for the robotics community and a challenge to be addressed. Natural environments are large scale and plain of 3D features. Conventional 2D metric maps are not appropriate to represent and support the navigation across such environments. The team has addressed and will continue to work on the identification of natural features, 3D semantic maps and visual odometry based on a Sparse Distributed Memory approach.
For some robotic tasks, especially those that are intrinsically distributed and complex (e.g. covering a wide area in a natural environment, surveillance of large infrastructures, search and rescue of victims in the aftermath of a catastrophic incident, etc.), a team of several cooperative mobile robots is required to either make viable the mission accomplishment or, at least, accomplish the mission with better performance than a single robot. Cooperative robotic systems (CRS) have received significant attention by the robotics community for the past two decades, because their successful deployment has unquestionable social and economic relevance in many application domains. Due to the expendability of individual robots, CRS may substitute people in risky and harsh scenarios. Furthermore, in less risky scenarios, CRS may still relieve people from collective tasks that are intrinsically monotonous and repetitive and allow them to be occupied by nobler tasks. The benefits of CRS hinge, however, on suitable internal team organization to achieve coordination and cooperation, which in turn requires efficient sharing of information among robotic teammates. One of the key scientific objectives is to attain such collective coordination in a distributed way, in order to avoid the existence of a central point of failure, and attain resilient robotic teams whose collective performance degrades gracefully upon the occurrence of robot’s individual failures. Effective coordination in distributed architectures allows reducing the gap between suboptimal performance inherent to their distributed nature (decisions have to be made with partial information) and potentially optimal performance of centralized approaches. In this regard, methods of multi-sensor fusion for cooperative perception, and methods of efficient information sharing, are required to attain high collective performance with distributed control. Orthogonally to deliberative cooperation, swarm robotic teams have been investigated to attain emergent cooperation and coordination through the interaction of very simplistic individual robot behaviors.
Inspection, maintenance and cleaning of 3D human made structures, including pipes, reservoirs, buildings, bridges, wind turbines, and ship hulls, is currently performed by human workers, and many of these works are categorized as Dirty, Difficult and Dangerous (DDD) works. It is desirable to substitute as much as possible such DDD works by automated robotics solutions. Along with the progresses in the individual technologies regarding the climbing and terrestrial machines, their automation algorithms such as localization and mapping and communication between agents, as well as the robotics operating systems, the strategic goal in this field is to join state of the art technologies in order to automate the whole inspection scenario as much as possible. Therefore the strategic goal in this regards aims at cooperative autonomous inspection of 3D human made structures with multiple agents, including climbing, flying, in-pipe and terrestrial robots. This includes 3D mapping of the structures to be inspected as well as the field where the structure is mounted and localization of all agents. Inspecting robots may cooperate to localize each other, share their map and act as beacons for each other. An example scenario mission is mapping of a 3D structure in a plant or field, inspection and cleaning of the whole structure, and localizing and reporting the spots with problems such as material degradation, cracks, etc.