The main bottleneck in evolutionary robotics has traditionally been the time required to evolve robot controllers. However with the continued acceleration in computational resources, the main bottleneck is now the tim...
详细信息
ISBN:
(纸本)9781450311786
The main bottleneck in evolutionary robotics has traditionally been the time required to evolve robot controllers. However with the continued acceleration in computational resources, the main bottleneck is now the time required for an investigator to create a robot simulator, a neural network, evolutionary algorithm and fitness function that together produce the desired behavior. Here we introduce a software framework that allows a user to conduct evolutionary robotics experiments without having to write any software themselves: the user defines the robot morphology, task environment and fitness function interactively;a neural network is constructed based on the robot's morphology;and an evolutionary algorithm optimizes desired behavior. We here show that this approach allows users to overcome one of the main limitations of evolutionary algorithms-recognizing and then preventing cntrapment in local optima-in a continuous, code free manner.
Most modern applications of evolutionary robotics (ER) rely upon computer-based physics simulations in order to model the behavior of the systems in question. One of the greatest challenges in the field of ER, therefo...
详细信息
ISBN:
(纸本)9781450349390
Most modern applications of evolutionary robotics (ER) rely upon computer-based physics simulations in order to model the behavior of the systems in question. One of the greatest challenges in the field of ER, therefore, is the development of robust, high-precision and accurate physics simulators that can model all necessary and relevant real-world interactions in an computationally efficient manner. Up until now, most popular ER simulators are nonetheless deficient in one or many of these properties. Here we introduce a new competitive simulator, the Baseline-Realistic Objective Open-Ended Kinematics Simulator (BROOKS) that outperforms other off-the-shelf simulators in most criteria. Our simulator is free, open-sourced, and easy to modify. It can model a wide range of robotic platforms, substrates and environments. Moreover, we claim solutions produced within the BROOKS simulator perform almost identically in the real-world, thereby helping to address one of the most challenging aspects of simulation in evolutionary robotics: the Reality Gap. Ultimately, we believe that BROOKS will establish a new baseline against which all other simulators should be compared.
In evolutionary robotics, robot controllers are often evolved in a separate development phase preceding actual deployment - we call this off-line evolution. In on-line evolutionary robotics, by contrast, robot control...
详细信息
ISBN:
(纸本)9781450305570
In evolutionary robotics, robot controllers are often evolved in a separate development phase preceding actual deployment - we call this off-line evolution. In on-line evolutionary robotics, by contrast, robot controllers adapt through evolution while the robots perform their proper tasks, not in a separate preliminary phase. In this case, individual robots can contain their own self-sufficient evolutionary algorithm (the encapsulated approach) where individuals are typically evaluated by means of a time sharing scheme: an individual is given the run of the robot for some amount of time and fitness corresponds to the robot's task performance in that period. Racing was originally introduced as a model selection procedure that quickly discards clearly inferior models. We propose and experimentally validate racing as a technique to cut short the evaluation of poor individuals before the regular evaluation period expires. This allows an increase of the number of individuals evaluated per time unit, but it also increases the robot's actual performance by virtue of abandoning controllers that perform inadequately. Our experiments show that racing can improve the performance of robots that adapt their controllers by means of an on-line evolutionary algorithm significantly.
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...
详细信息
ISBN:
(纸本)9781479944941
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.
The aim of evolutionary robotics is to develop neural systems for behavior control of autonomous robots. For non-trivial behaviors or non-trivial machines the implementation effort for suitably specialized simulators ...
详细信息
ISBN:
(纸本)9783642173189
The aim of evolutionary robotics is to develop neural systems for behavior control of autonomous robots. For non-trivial behaviors or non-trivial machines the implementation effort for suitably specialized simulators and evolution environments is often very high. The Neurodynamics and evolutionary robotics Development Kit (NERD), presented in this article, is a free open-source framework to rapidly implement such applications. It includes separate libraries (1) for the simulation of arbitrary robots in dynamic environments, allowing the exchange of underlying physics engines, (2) the simulation, manipulation and analysis of recurrent neural networks for behavior control, and (3) an extensible evolution framework with a number of neuro-evolution algorithms. NERD comes with a set of applications that can be used directly for many evolutionary robotics experiments. Simulation scenarios and specific extensions can be defined via XML, scripts and custom plug-ins. The NERD kit is available at *** under the GPL license.
The self-organizing bio-hybrid collaboration of robots and natural plants allows for a variety of interesting applications. As an example we investigate how robots can be used to control the growth and motion of a nat...
详细信息
ISBN:
(纸本)9781509035342
The self-organizing bio-hybrid collaboration of robots and natural plants allows for a variety of interesting applications. As an example we investigate how robots can be used to control the growth and motion of a natural plant, using LEDs to provide stimuli. We follow an evolutionary robotics approach where task performance is determined by monitoring the plant's reaction. First, we do initial plant experiments with simple, predetermined controllers. Then we use image sampling data as a model of the dynamics of the plant tip xy position. Second, we use this approach to evolve robot controllers in simulation. The task is to make the plant approach three predetermined, distinct points in an xy-plane. Finally, we test the evolved controllers in real plant experiments and find that we cross the reality gap successfully. We shortly describe how we have extended from plant tip to many points on the plant, for a model of the plant stem dynamics. Future work will extend to two-axes image sampling for a 3-d approach.
evolutionary robotics simulations can serve as a tool to clarify counterintuitive or dynamically complex aspects of sensorimotor behaviour. We present a series of simulations that has been conducted in order to aid th...
详细信息
ISBN:
(纸本)9783540749127
evolutionary robotics simulations can serve as a tool to clarify counterintuitive or dynamically complex aspects of sensorimotor behaviour. We present a series of simulations that has been conducted in order to aid the interpretation of ambiguous empirical data on human adaptation to delayed tactile feedback. Agents have been evolved to catch objects falling at different velocities to investigate the behavioural impact that lengthening or shortening of sensory delays has on the strategies evolved. A detailed analysis of the evolved model agents leads to a number of hypotheses for the quantification of the existing data, as well as to ideas for possible further empirical experiments. This study confirms the utility of evolutionary robotics simulation in this kind of interdisciplinary endeavour.
Mainstream approaches to modelling cognitive processes have typically focused on (1) reproducing their neural underpinning, without regard to sensory-motor systems and (2) producing a single, ideal computational model...
详细信息
Mainstream approaches to modelling cognitive processes have typically focused on (1) reproducing their neural underpinning, without regard to sensory-motor systems and (2) producing a single, ideal computational model. evolutionary robotics is an alternative possibility to bridge the gap between neural substrate and behavior by means of a sensory-motor apparatus, and a powerful tool to build a population of individuals rather than a single model. We trained 4 populations of neurorobots, equipped with a pan/tilt/zoom camera, and provided with different types of motor control in order to perform a cancellation task, often used to tap spatial cognition. Neurorobots’ eye movements were controlled by (a) position, (b) velocity, (c) simulated muscles and (d) simulated muscles with fixed level of zoom. Neurorobots provided with muscle and velocity control showed better performances than those controlled in position. This is an interesting result since muscle control can be considered a particular type of position control. Finally, neurorobots provided with muscle control and zoom outperformed those without zooming ability.
This workshop presentation describes the general concepts behind embodied evolution, and intends to provide an up-to-date view of lessons learned and current open issues.
ISBN:
(纸本)9781450334884
This workshop presentation describes the general concepts behind embodied evolution, and intends to provide an up-to-date view of lessons learned and current open issues.
In this work a new method to evolutionary robotics is proposed, it combines into asingle framework, learning from reality and simulations. An illusory sub-system is incorporated as an integral part of an autonomous sy...
详细信息
In this work a new method to evolutionary robotics is proposed, it combines into asingle framework, learning from reality and simulations. An illusory sub-system is incorporated as an integral part of an autonomous system. The adaptation of the illusory system results from minimizing differences of robot behavior evaluations in reality and in simulations. Behavior guides the illusory adaptation by sampling task-relevant instances of the world. Thus explicit calibration is not required. We remark two attributes of the presented methodology: (i) it is a promising approach for crossing the reality-gap among simulation and reality in evolutionary robotics, and (ii) it allows to generate automatically models and theories of the real robot environment expressed as simulations. We present validation experiments on locomotive behavior acquisition for legged robots.
暂无评论