Abstract:
Robots have been successfully deployed during natural disasters to perform remote
search and rescue missions. These robots are tasked under human operator
supervision. For remote operations, network connectivity is essential. However,
there is a scarcity of network infrastructure in a post-disaster scenario and the
robots may lose connectivity with the operator. If the environment is dynamic, a
robot may be disconnected from the base station, in which case, either it stays at
the current location until network is re-established or searches in the environment
to re-establish the connection. An intuitive mechanism of returning to the last
network connected location may be an inefficient strategy. In this thesis, we use
foraging concepts from the animal kingdom to address the problem of connection
re-establishment in sparse network coverage scenarios. We use a combination of
L´evy walks, past path memory and convex hull concepts to develop an efficient
hybrid model that allows the robots to escape from no-network areas. Simulation
results are presented that show the superiority of our hybrid model in establishing
connectivity with the base station compared to L´evy only search, memory-based
search and random search.
It is also important to ensure that the time spent on interaction between the human
operator and the robots/agents is as minimal as possible. If the operator has to
control a large number of agents, known as a swarm, then it becomes time consuming
for him to interact with each agent. The interaction time can be reduced if
the operator controls a subset of agents to guide the behaviour of the swarm. This
can be even further reduced by removing operator control over selection of agents
within a swarm. Hence, we examine how automatic selection of agents, within a
swarm, should be done so as to influence the other agents to complete the assigned
task. We also find out how many influencing agents should be selected and where
they should be located for efficient relocation of the swarm without any swarm
fragmentation.
