Gradient estimating and climbing for source searching by multi-agent system

A new method is proposed that multi-agent system estimates the gradient of a scalar diffusion field and advances towards the diffusion source. The method in this paper integrates three aspects: formation control, gradient estimation and the corresponding gradient climbing control. Formation control guarantees that motes are in relative right positions, configuring a suitable sensor array, so that their measurements (maybe temperature) generated by a diffusion source (maybe a fire source) can be efficiently used by the gradient climbing control. The gradient climbing control law is then superposed to every mote, through which the mote network estimates the source field's gradient and climbs the gradient until the diffusion source is reached. The errors and stability are analyzed and the results indicate almost global convergence properties of this new method. Finally this method is extended to the scalar fields which have several extrema but only one maximum value.