Högskolan i Skövde

A growing body of research within the field of intelligent robotics argues for a view of intelligence drastically different from classical artificial intelligence and cognitive science. The holistic and embodied ideas expressed by this research promote the view that intelligence is an emergent phenomenon. Similar perspectives, where numerous interactions within the system lead to emergent properties and cognitive abilities beyond that of the individual parts, can be found within many scientific fields. With the goal of understanding how behavior may be represented in robots, the present review tries to grasp what this notion of emergence really means and compare it with a selection of theories developed for analysis of human cognition, including the extended mind, distributed cognition and situated action. These theories reveal a view of intelligence where common notions of objects, goals, language and reasoning have to be rethought. A view where behavior, as well as the agent as such, is defined by the observer rather than given by their nature. Structures in the environment emerge by interaction rather than recognized. In such a view, the fundamental question is how emergent systems appear and develop, and how they may be controlled.

One common approach to the robot learning technique Learning From Demonstration, is to use a set of pre-programmed skills as building blocks for more complex tasks. One important part of this approach is recognition of these skills in a demonstration comprising a stream of sensor and actuator data. In this paper, three novel techniques for behavior recognition are presented and compared. The first technique is function-oriented and compares actions for similar inputs. The second technique is based on auto-associative neural networks and compares reconstruction errors in sensory-motor space. The third technique is based on S-Learning and compares sequences of patterns in sensory-motor space. All three techniques compute an activity level which can be seen as an alternative to a pure classification approach. Performed tests show how the former approach allows a more informative interpretation of a demonstration, by not determining "correct" behaviors but rather a number of alternative interpretations.

The paper describes and formalizes the concepts and assumptions involved in Learning from Demonstration (LFD), a common learning technique used in robotics. LFD-related concepts like goal, generalization, and repetition are here defined, analyzed, and put into context. Robot behaviors are described in terms of trajectories through information spaces and learning is formulated as mappings between some of these spaces. Finally, behavior primitives are introduced as one example of good bias in learning, dividing the learning process into the three stages of behavior segmentation, behavior recognition, and behavior coordination. The formalism is exemplified through a sequence learning task where a robot equipped with a gripper arm is to move objects to specific areas. The introduced concepts are illustrated with special focus on how bias of various kinds can be used to enable learning from a single demonstration, and how ambiguities in demonstrations can be identified and handled.

A model-free learning algorithm called Predictive Sequence Learning (PSL) is presented and evaluated in a robot Learning from Demonstration (LFD) setting. PSL is inspired by several functional models of the brain. It constructs sequences of predictable sensory-motor patterns, without relying on predefined higher-level concepts. The algorithm is demonstrated on a Khepera II robot in four different tasks. During training, PSL generates a hypothesis library from demonstrated data. The library is then used to control the robot by continually predicting the next action, based on the sequence of passed sensor and motor events. In this way, the robot reproduces the demonstrated behavior. PSL is able to successfully learn and repeat three elementary tasks, but is unable to repeat a fourth, composed behavior. The results indicate that PSL is suitable for learning problems up to a certain complexity, while higher level coordination is required for learning more complex behaviors.

Two methods for behavior recognition are presented and evaluated. Both methods are based on the dynamic temporal difference algorithm Predictive Sequence Learning (PSL) which has previously been proposed as a learning algorithm for robot control. One strength of the proposed recognition methods is that the model PSL builds to recognize behaviors is identical to that used for control, implying that the controller (inverse model) and the recognition algorithm (forward model) can be implemented as two aspects of the same model. The two proposed methods, PSLE-Comparison and PSLH-Comparison, are evaluated in a Learning from Demonstration setting, where each algorithm should recognize a known skill in a demonstration performed via teleoperation. PSLH-Comparison produced the smallest recognition error. The results indicate that PSLH-Comparison could be a suitable algorithm for integration in a hierarchical control system consistent with recent models of human perception and motor control.

In this chapter, the prediction algorithm Predictive Sequence Learning (PSL) is presented and evaluated in a robot Learning from Demonstration (LFD) setting. PSL generates hypotheses from a sequence of sensory-motor events. Generated hypotheses can be used as a semi-reactive controller for robots. PSL has previously been used as a method for LFD, but suffered from combinatorial explosion when applied to data with many dimensions, such as high dimensional sensor and motor data. A new version of PSL, referred to as Fuzzy Predictive Sequence Learning (FPSL), is presented and evaluated in this chapter. FPSL is implemented as a Fuzzy Logic rule base and works on a continuous state space, in contrast to the discrete state space used in the original design of PSL. The evaluation of FPSL shows a significant performance improvement in comparison to the discrete version of the algorithm. Applied to an LFD task in a simulated apartment environment, the robot is able to learn to navigate to a specific location, starting from an unknown position in the apartment.

A method for Learning from Demonstration (LFD) is presented and evaluated on a simulated Robosoft Kompai robot. The presented algorithm, called Predictive Sequence Learning (PSL), builds fuzzy rules describing temporal relations between sensory-motor events recorded while a human operator is tele-operating the robot. The generated rule base can be used to control the robot and to predict expected sensor events in response to executed actions. The rule base can be trained under different contexts, represented as fuzzy sets. In the present work, contexts are used to represent different behaviors. Several behaviors can in this way be stored in the same rule base and partly share information. The context that best matches present circumstances can be identified using the predictive model and the robot can in this way automatically identify the most suitable behavior for precent circumstances. The performance of PSL as a method for LFD is evaluated with, and without, contextual information. The results indicate that PSL without contexts can learn and reproduce simple behaviors. The system also successfully identifies the most suitable context in almost all test cases. The robot's ability to reproduce more complex behaviors, with partly overlapping and conflicting information, significantly increases with the use of contexts. The results support a further development of PSL as a component of a dynamic hierarchical system performing control and predictions on several levels of abstraction.