Evolution of neural controllers for competitive game playing with teams of mobile robots

Nelson, Andrew; Grant, Edward; Henderson, Thomas C.

doi:10.1016/j.robot.2004.01.001

Cited by 47 publications

(47 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This allows a single progressive aggregate fitness function based only on number of wins in a tournament to be used to drive selection. This concept has received some attention in ER in relation to evolving populations of controllers for competitive behaviors [10,23].…”

Section: B State Of the Art Methods In Ermentioning

confidence: 99%

“…In the following decade, many of these tasks were revisited and robots were evolved using much less a priori information in the selection process. During this time period a series of more elaborate behaviors were evolved in larger environments, using color vision for sensing or requiring greater use of memory (see for example [23,56,57]). There were also advancements in methodologies, including refinement of real-robot coupled simulators, evolutionary neural networks, and coevolution of morphology and control [54].…”

Section: Background and Historymentioning

confidence: 99%

“…There are several variants on this basic theme. In [23] teams of robots competed in the same extended maze arena to locate goals differentially and in [57] a complex environment-mediated differential homing/repulsion task was studied (see Section VI.A).…”

Section: A Target Homingmentioning

confidence: 99%

“…However, many of these tasks have been shown directly to have relatively simple solutions. A common component of many ER experiments is to hand code controllers for comparison [23,59,77]. For the vast majority of tasks investigated in the literature, such hand coded controllers are in fact extremely simple, requiring only a few pages of code, yet are shown to be competitive with evolved controllers in most cases.…”

Section: A Benchmark Er Tasks Don't Represent the General Casementioning

confidence: 99%

“…But this raises another difficulty: even with a well-defined goal, formulating a goal-oriented fitness function capable of traversing a vast ultrahigh-dimensional search space is likely to be intractable for complex tasks without the aid of detailed domain knowledge [21]. Functions that aggregate fitness evaluation based on success or failure to complete a task suffer from the inability to generate any fitness signal in randomly initiated populations [22,23]. For goal-driven evolutionary searches for behaviors of sufficient complexity, fitness functions that do not encode significant features of a particular solution will tend to define fitness landscapes in which regions outside a given radius of any fitness optimum consist primarily of plateaus with no detectable gradient (the Bootstrap Problem).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Embodied artificial life at an impasse can evolutionary robotics methods be scaled?

Nelson¹

2014

2014 IEEE Symposium on Evolving and Autonomous Learning Systems (EALS)

View full text Add to dashboard Cite

Abstract-Evolutionary robotics (ER) investigates the application of artificial evolution toward the synthesis of robots capable of performing autonomous behaviors. Over the last 25 years, researchers have reported increasingly complex evolved behaviors, and have compiled a de facto set of benchmark tasks. Perhaps the best known of these is the obstacle avoidance and target homing task performed by differential drive robots. More complex tasks studied in recent ER work include augmented variants of the rodent T-maze and complex foraging tasks. But can proof-of-concept results such as these be extended to evolve complex autonomous behaviors in a general sense? In this topical analysis paper we survey relevant research and make the case that common tasks used to demonstrate the effectiveness of evolutionary robotics are not characteristic of more general cases and in fact do not fully prove the concept that artificial evolution can be used to evolve sophisticated autonomous agent behaviors. Robots capable of performing many of the tasks studied in ER have now been evolved using nearly aggregate binary success/fail fitness functions. However, arguments used to support the necessity of incremental methods for complex tasks are essentially sound. This raises the possibility that the tasks themselves allow for relatively simple solutions, or span a relatively small candidate solution set. This paper presents these arguments in detail and concludes with a discussion of current ER research.

show abstract

Section: B State Of the Art Methods In Ermentioning

confidence: 99%

Section: Background and Historymentioning

confidence: 99%

Section: A Target Homingmentioning

confidence: 99%

Section: A Benchmark Er Tasks Don't Represent the General Casementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Embodied artificial life at an impasse can evolutionary robotics methods be scaled?

Nelson¹

2014

2014 IEEE Symposium on Evolving and Autonomous Learning Systems (EALS)

View full text Add to dashboard Cite

show abstract

Data Communication Assistance via Swarm Robotic: A Behaviour Creation Comparison

Smith

Khan

Aleti

et al. 2020

Lecture Notes in Computer Science

View full text Add to dashboard Cite

This paper compares two behaviour creation algorithms for Unmanned Aerial Vehicle swarms. The objective of this work is to have a self-organising robotic swarm which unintrusively assists humans in network restricted environments by facilitating data-transfer between disconnected groups. The behaviour creation algorithms explored are a well-cited approach for evolving neural networks (ENN) and our Learning Classifier System (LCS) approach. We utilise simulations in randomised, obstacle dense landscapes to validate, compare and contrast these two algorithms. We explore the algorithms both with and without the environment layout being known to the swarm a priori. These simulations show the ENN algorithm struggles to create appropriate behaviours for this complex data-transfer task, resulting in an unassertive swarm. Our LCS behaviours have the swarm operate with up to 61% of the fitness range above ENN in both expected and unexpected environments, resulting in the desired human assistance.

show abstract

Evolution, Self-organization and Swarm Robotics

Trianni

Nolfi

Dorigo

Natural Computing Series

View full text Add to dashboard Cite

Summary. The activities of social insects are often based on a self-organising process, that is, "a process in which pattern at the global level of a system emerges solely from numerous interactions among the lower-level components of the system"(see [4], p. 8). In a self-organising system such as an ant colony, there is neither a leader that drives the activities of the group, nor are the individual ants informed about a global recipe or blueprint to be executed. On the contrary, each single ant acts autonomously following simple rules and locally interacting with the other ants. As a consequence of the numerous interactions among individuals, a coherent behaviour can be observed at the colony level.A similar organisational structure is definitely beneficial for a swarm of autonomous robots. In fact, a coherent group behaviour can be obtained providing each robot with simple individual rules. Moreover, the features that characterise a self-organising system-such as decentralisation, flexibility and robustness-are highly desirable also for a swarm of autonomous robots. The main problem that has to be faced in the design of a self-organising robotic system is the definition of the individual rules that lead to the desired collective behaviour. The solution we propose to this design problem relies on artificial evolution as the main tool for the synthesis of self-organising behaviours. In this chapter, we provide an overview of successful applications of evolutionary techniques to the evolution of self-organising behaviours for a group of simulated autonomous robots. The obtained results show that the methodology is viable, and that it produces behaviours that are efficient, scalable and robust enough to be tested in reality on a physical robotic platform.

show abstract

Evolution of neural controllers for competitive game playing with teams of mobile robots

Cited by 47 publications

References 23 publications

Embodied artificial life at an impasse can evolutionary robotics methods be scaled?

Embodied artificial life at an impasse can evolutionary robotics methods be scaled?

Data Communication Assistance via Swarm Robotic: A Behaviour Creation Comparison

Evolution, Self-organization and Swarm Robotics

Contact Info

Product

Resources

About