Key Aspects of PSO-Type Swarm Robotic Search: Signals Fusion and Path Planning

Extending the particle swarm optimization (PSO) algorithm to be one of systemic modeling and controlling tools, several research groups investigate target search with swarm robots (simulated or physical) respectively (Doctor et al., 2004; Hereford & Siebold, 2008; Jatmiko et al., 2007; Marques et al., 2006; Pugh & Martinoli, 2007; Xue & Zeng, 2008). The common idea they hold is to map such swarm robotic search to PSO and deal it with by employing the existing bio-inspired approaches to the latter case in a similar way (Xue et al., 2009). Of the mapping relations, some aspects including fitness evaluate and path planning have to be especially considered because PSO-type algorithm working depends heavily upon them. Unlike regarding these respects in PSO, however, the actual characteristics of robot and complexity of sensing to environment make it impossible to be simplified even ignored. Bear that in mind, we might as well explore some representative research work. Pugh et al. compare the similarities and differences in properties between real robot and ideal particle, then extend PSO directly to model multiple robots for studying at an abstract level the effects of changing parameters of the swarm system (Pugh & Martinoli, 2007). Xue et al. simplify characteristics of robot by treating each physical robot as a first order inertial element to study mechanism of limited sensing and local interactions in swarm robotic search (Xue & Zeng, 2008). Doctor et al. discuss applying PSO for multiple robot searches, whose focus is on optimizing the parameters of their algorithm (Doctor et al., 2004). Jatmiko et al. exert mobile robots for plume detection and traversal, with utilizing a modified form of PSO to control the robots and consider how the robots respond to search space changes such as turbulence and wind changes (Jatmiko et al., 2007). Hereford et al. consider how well the PSO-based robot search will scale to large numbers of robots by designing specific communication strategies. Based upon this, they have published results of implementing their PSO variants in actual hardware robot swarms (Hereford & Siebold, 2008). Marques et al. analytically compare PSO-based cooperative search and gradient search as well as biased-random walk search to try to find out which performing well in search efficiency. Due to the exchange of information between neighbors in the first search mode, PSO-type olfactory guided search possesses merit in search properties over its two competitors (Marques et al., 2006). It is clear that all of works mentioned above neither involve target search with PSO-type control algorithm under conditions of realistic sensing to environment, nor handle the problem of obstacle avoidance in the process of target search. On the contrary, each of them assumes a potential target in search space to give off a diffuse residue that can be detected by a single 4