1 Using Physics-Like Interaction Law to perform Active Environment Recognition in Mobile Robotics A. Hazan, F. Davesne, V. Vigneron, H. Maaref Laboratoire Syst` emes Complexes (LSC), CNRS FRE 2494 CE 1433 Courcouronnes 40, Rue du Pelvoux 91020 Evry Cedex Phone: +0033169477504, E-mail: { hazan,davesne,vvigne,maaref } @iup.univ-evry.fr point of the sensorimotor hypothesis for biological entities Abstract - In this article, we give some insights of a novel method for active environment recognition in mobile robotics. [HEL 21], [GIB 79]. Loops including motor neurons and The basic idea consists on utilizing a Physics-like interac- neurons associated with senses have been discovered in the tion law to fix a relation between sensors and effectors val- brain. Moreover, it has been shown that eye saccades are ues at any time. Our main assumption is that the trajectory necessary in the human recognition process and that ac- of the robot in the phase space, which depends uniquely on tive movement may help people to disambiguate artificial its environment -when the law and the nature of the robot scenes [WEX 01]. This precise idea has been exploited in are fixed- may discriminate environments better than classi- robotics, in the field of active vision [BAL 90], [BAJ 88]. cal Data Analysis Approaches (DAA). In order to test our as- sumption, we choose to model an analogical robot which light sensor amplitudes and wheels speed are coupled in a set of Recently, our team have also done some work on the use differential equations. As a result, we show that our Interac- of fractal dimension to caracterize the specificity of data tionist Approach (IA) is tractable and perform well for dis- gathered by a moving mobile robot [VIG 02]. This work criminating simple environments, comparing to a data analy- has led to the conclusion that a disconnection between the sis (DA) strategy. strategy of movement and the data analysis process may carry poor results. Keywords — Mobile Robotics, Dynamic Systems, Environe- ment Recognition, Physics-like interaction (a) I. I NTRODUCTION A. Framework Value function According to the traditional point of view in mobile Physical Data Acquisition data set robotics, sensing is a passive process (i.e. gather data us- Data Analysis Output World Process ing sensors) whereas moving in the world is an active one. Any task of the robot fulfills the following straight forward Machine schema: sense the world → analyze gathered data → act in (b) the world (see Fig. 1, (a)). The Data Acquisition (DAQ) process relies on a set of sensors that transduct and dig- Interaction Value function itize some environmental variables. The result is called Constraint a set of data . It feeds in a straightforward manner a DA Physical Data Acquisition data set stage, which aim is to make the data set useful to the ex- Data Analysis Output World Process perimenter, with respect to a value function that embodies the experimenters’ needs to understand the physical world. Machine Acting as best as possible (to achieve a precise goal) im- Fig. 1. (a) Classical approach. (b) Interactionist approach. plies a DA process ruled by an optimal (or suboptimal) de- cision making policy that leads to an optimal (or subopti- mal) action of the robot in the world. The (statistical) preci- B. Main assumption sion and reliability of the resulting task depends mainly on the way it has been modeled and on the data analysis pro- The general idea implies that the acquisition process is cess, because the DAQ process is considered to be fixed. made of two interconnected modules: sensing and acting. However, our assumption is much more precise than that. It Our general claim supposes that the reliability or the pre- relies on the existence of a physics-like interaction law that cision of the former results may be enhanced by consider- links sensors and effectors values (see Fig. 1, (b)). This ing active data acquisition processes. In the case of mo- assumption has immediate consequences: sensory and mo- bile robotics, that involves considering ”small” and ”fast” tor variables are instantly codetermined, with no possibil- movements of the robot performed during the data acqui- ity to orientate that link, e.g. to say that if sensor values sition time-lapse. The idea that movement is crucial to are changed by a given amount, then effectors values will gather ”good” data is not new. Historically, it is the key change in a certain way. In the case we depict in this paper,
the sensory motor law is modeled by a set of coupled differ- robot first evolves locally in a given target environment, ential equations. The solutions (when then exist) are trajec- following some trajectory we will discuss later in this pa- per. Then, it is presented to a series of k distractive envi- tories in the phase space (combining sensors and effectors variables). The class of solutions may be interpreted as the ronments (i.e. that may or may not correspond to the target environment) where it evolves during d as in the first stage. set of all possible behaviors of the robot facing all possible worlds. A particular trajectory in the phase space is deter- The aim of this task consists on identifying the target en- mined during the experiment , when the robot is facing a vironment among the distractive environments on the basis particular world. Thus, two different trajectories (given a of the data collected and analyzed during the various ex- certain distance) may be associated to two different envi- periments. ronments (see Fig. 2) : this determines the basic principle for an environment recognition process . In the classical approach, called DA Approach (DAA) , this implies to put the robot in two given environments, to ex- ecute the same trajectory in both cases and to compare the X 1 corresponding sensor values. The discrimination between two environments is then given by a distance between two sensor data vectors. In our approach, called Interactionist Approach (IA) , the robot moves in order to fulfill the interaction law. And X 2 its motion during the time-lapse d depends on the sensor values, hence the environment. So, one cannot force the robot’s trajectory over d , but one may hope that this tra- jectory is a signature of a local robot/environment interac- tion. The discrimination between two environments is then given by a distance between two trajectories in the phase possible environments space combining sensor and effector variables. Fig. 2. Different trajectories associated to different environments. X 1 and X 2 are the variables of the phase space. We depict D. Issues covered by this paper two planes over the environment axis, representing two dif- ferent environments. The different trajectories in the same We have chosen to model the interaction law with a set of plane represent different behaviors of the robot placed in dis- tinct areas of the same environment. coupled differential equations, which is a particular way of implementing our assumption. In this paper, we detail The existence of a physics-like interaction law is a strong issues arising from this choice and provide simple exam- constraint because the relation between sensory motor vari- ples in which artificial worlds may be discriminated by a ables must be fulfilled at any time . It is based on an ac- simulated robot after using our approach and compared to tion/reaction procedure: the world (which is a priori un- the classical approach. These results are the beginning of a known) acts on the robot by the way of the sensor values work leading to an extended comparison, both theoretical and the movement of the robot and, at the same time, the and experimental, between the results obtained by the clas- machine reacts to adapt its internal parameters (which are sical approach and ours. known) in order to follow the interaction law. This ac- tion/reaction procedure has already been successfully uti- II. M ODELIZATION OF OUR ASSUMPTION lized to design a reinforcement learning algorithm onto which convergence proofs may easily be given [DAV 99], A. General Principle for discriminating environments in [DAV 04]. One particular advantage consists on the possi- the IA approach bility to determine the class of solutions before the exper- In IA, we impose a local discrimination criterion on the iment . In our case, this permits to have an idea about the Data Acquisition stage. To make it clearer, we assume that similarity of the shapes of the possible trajectories. The the Data Acquisition step defines a multidimensional phase more two shapes are ”different”, the easiest it is to discrim- space that includes both motor and sensitive data, in which inate the two associated environments. a state is called X . The interaction of the robot is thus rep- resented in this space by a trajectory T : t → X ( t ) . We then suppose that the robot can take two simultaneous dif- C. Environment recognition - interactionist versus classi- ferent - but “close” - interactive measures, i.e. that it can cal approach follow at the same time two different but close trajectories T 1 and T 2 of the phase portrait. We also assume that the in- In this article we focus on an environment recognition task in which recognition is made by gathering data over a teraction is not a completely deterministic process, but that fixed time-lapse d . In a general sense, that means that the it has a stochastic component. We may then think of two
Recommend
More recommend
