Patrick Ediger

A scalable configurable architecture for the massively parallel GCA model

The global cellular automata model (GCA) is a massively parallel computation model which extends the classical cellular automata model (CA) with dynamic global neighbors. We present for that model a data parallel architecture which is scalable in the number of parallel pipelines and which uses application specific operators (adapted operators). The instruction set consists of control and RULE instructions. A RULE computes the next cell contents for each cell in the destination object. The machine consists of P pipelines. Each pipeline has an associated primary memory bank and has access to the global memory (real or emulated multiport memory). The diffusion of particles was used as an example in order to demonstrate the adaptive operators, the machine programming and its performance. Particles which point to each other within a defined neighborhood search space are interchanged. The pointers are modified in each generation by apseudo random function. The machine with up to 32 pipelines was synthesized for an Altera FPGA for that application.

Routing in the triangular grid with evolved agents

Given a triangular grid of N cells (communication nodes) with toroidal connections. The goal was to solve the routing problem with N/2 agents, each of the agents having the task to a transport a message from a source to a target. This task is also known as multiple target searching. The agents shall behave according to a control algorithm implemented as finite state machine (FSM). Using a genetic procedure (island genetic algorithm) algorithms were evolved that could solve successfully all the test cases under consideration. For comparison, intelligent random walkers were defined, which directly try to move to the target, or deviate from their way with a certain probability. It turned out that the evolved agents perform the task 22% faster than the intelligent random walkers.

Evolving Multi-creature Systems for All-to-All Communication

Several creatures are moving around in a cellular automata grid. At a certain point of time all creatures want to exchange their information with all others (all-to-all communication). The goal is to find an optimal rule for the movement of the creatures in order to exchange their information as fast as possible. The information exchange is only possible when the creatures meet each other and when they form certain defined local patterns (communication situations). Possible communication situations are exemplarily shown in Fig. 1. In the cases a, b, c the creatures are directly in contact. But it is a matter of definition whether such situations allow communication. For this investigation we have defined the communication patterns d, e, f. A reason could be that communication can only take place if a mediator/negotiator is used between them. Furthermore the mediator may perform a particular computation (e. g., average, maximum, priority select). Such conflicts occur when creatures want to move to the same target position, like vehicles which are meeting in a cross-way. The center of the crossing can be interpreted as the mediator.

On the Effectiveness of Evolution Compared to Time-Consuming Full Search of Optimal 6-State Automata

The Creatures Exploration Problem is defined for an independent agent on regular grids. This agent shall visit all non-blocked cells in the grid autonomously in shortest time. Such a creature is defined by a specific finite state machine. Literature shows that the optimal 6-state automaton has already been found by simulating all possible automata. This paper tries to answer the question if it is possible to find good or optimal automata by using evolution instead of time-consuming full simulation. We show that it is possible to achieve 80% to 90% of the quality of the best automata with evolution in much shorter time.

Improving the Behavior of Creatures by Time-Shuffling

The goal is to optimize the behavior of moving creatures by using “time-shuffling” techniques. The “creatures’ exploration problem” is used as an example for a multi-agent problem modeled by cellular automata. The task of the creatures is to visit all empty cells in an environment with a minimum number of steps. The behavior of a creature is modeled by an automaton taking care of the collisions. Time-shuffling means that two behaviors (algorithms X and Y) are sequentially alternated with a certain time period. Ten different “uniform” (non-time-shuffled) algorithms with good performance from former investigations were used. We defined three time-shuffling modes differing in the way how the algorithms are interchanged. New metrics are used for such multi-agent systems, especially the success rate (number of successful explored environments) and the mean normalized work (cost). Time-shuffled systems with a time period of around 100 have resulted in much better success rates and lower cost compared to the uniform systems.

Rectangular vs triangular routing with evolved agents

A multiple target searching with evolved agents is performed in a cellular automata network to solve the routing problem in the square toroidal grid. The agents shall behave according to a control algorithm implemented as finite state machine (FSM). Using a genetic procedure, algorithms are evolved that could solve successfully all the training cases under consideration. In order to avoid deadlocks, a certain amount of randomness is added to the FSM. Intelligent random walkers (IW) are also considered. This paper is a companion paper on a previous work dealing with a similar protocol running in the triangular torus. It yields comparative performance results between rectangular and triangular routing, giving advantage to the latter.

Evolving hybrid time-shuffled behavior of agents

We searched for methods to evolve the hybrid behavior of moving agents for the All-to-All Communication task. The multi-agent system is modeled in two-dimensional Cellular Automata. An agent is controlled by one or more finite state machines. We use a time-shuffling method to join the state machines into one hybrid “algorithm”. We propose a method to directly evolve a hybrid behavior consisting of multiple state machines including their time-shuffling periods. We compared the evolved hybrid algorithms to other evolved non-hybrid algorithms (consisting of only one finite state machine) and to hybrid algorithms that were composed of separately evolved non-hybrid algorithms. The performance of the directly evolved hybrid algorithms was significantly better, and the computation time for the evolution was roughly the same.

All-to-all communication with CA agents by active coloring and acknowledging

We modeled a multi-agent system as a two-dimensional Cellular Automata and searched for a rule in order to solve the all-to-all communication task in shortest time. The rule contains two finite state machines (FSM) controlling the behavior of the uniform agents. The moving FSM controls the moving actions and the color FSM controls the changing of the cells color. Colors are used for indirect communication. In addition the agents receive an acknowledgment whenever they meet and communicate successfully. The FSMs were evolved by a genetic algorithm. It could be shown that acknowledging and especially coloring increases the performance of the agents. Certain initial configurations cannot be solved without coloring. Even with coloring, symmetric configurations cannot be solved when the initial colors are the same.