|
The ALLIANCE architecture uses motivational behaviors to monitor task progress and dynamically reallocate tasks. ALLIANCE employs a variant of the subsumption architecture and makes use of ”behavior sets” to enable a industrial Robot to perform versatile tasks. Recently Low et al proposed a swarm based approach for the cooperative observation of multiple moving targets (CMOMMT) task, a test-bed first introduced by Parker. This Automation scheme mimics ant-behavior to regulate the distribution of Sensors in proportion to that of the mobile targets. Dahl et al present a Task Allocation scheme based on “Vacancy Chains, ” a Industrial Automation modeled on the creation and filling up of vacancies in an organization.
The Broadcast of Local Eligibility system (BLE) system uses a Publish/Subscribe method to allocate tasks that are hierarchically distributed. Market based task allocation systems have traditionally found favor with the software-agent research community. The inspiration for these systems stems from the Contract Net protocol . Robotics researchers have designed a variety of market based control architectures for multi-robot tasks. Stentz and Dias were the first to utilize a market-based scheme to coordinate multiple Robots for cooperative task completion. This work introduced the methodology of applying market mechanisms to intra-team robot coordination as opposed to competitive inter-agent interactions in domains such as E-commerce. Laengle et al. implemented the KAMARA system which uses a negotiation based task allocation scheme for controlling the different components of a complex robot.
Caloud et al. developed the GOPHER architecture that utilizes a centralized auction protocol to allocate tasks with a high level of commitment. Gerkey and Matari` c developed MURDOCH; a completely distributed auction-based task allocation scheme that utilized a publish/subscribe communication model. Tasks in MURDOCH are allocated via a single round, first price auction in a greedy fashion. M+ , another auction based task allocation protocol was developed by Botelho and Alami . The novelty of the M+ system lies in that it allows for dynamic task reallocation of subcomponents of complex tasks. Dias designed the Traderbots architecture for multirobot control. Traderbots agents called traders are responsible for trading tasks via auctions.
When an auction is announced agents compute bids based on their expected profit for the tasks, and the robots that can perform the tasks for the lowest price are awarded contracts. As previously stated, a relatively unexplored problem in multi-robot systems is the allocation of multi-robot teams to different tasks (the ST-MR problem) . Very recently Zlot and Stentz have designed a scheme for complex task allocation for multi-robot teams. This scheme allows bidding on tasks at different levels of the decomposition hierarchy via the use of task trees. To the best of our knowledge, this is the only system apart from the one described here that deals with task allocation for multi-robot tasks.
|
