Styrning av autonoma system

Projektet genomförs på avdelningen för optimeringslära och systemteori, KTH.

Gruppstorlek: 2-6

Speciella förkunskapskrav:

Grundkurserna i matematik och optimeringslära. Kunskaper i programmering i Matlab eller C++. Kunskap i matematisk systemteori är ett plus.

Projektansvarig:: Professor Xiaoming Hu, hu@kth.se

Handledare: Xiaoming Hu, Johan Thunberg, Johan Markdahl.

Projektbeskrivning

Student kan välja ett från de två följande projekten.

Project 1: Pursuit Evasion Problems in Polygonal Environments

In the visibility based pursuit evasion problem, a group of pursuers shall find a search strategy such that an evader moving arbitrarily fast and with unknown start location will be captured (or seen) no matter what path it decides to take. The pursuers and the evader are modeled as points in a connected bounded region in the plane, containing polygonal holes. Two points are visible from each other if the line between them does not intersect any polygonal hole. An obvious application of the pursuit evasion problem is where security guards or robots are to clear an office, a warehouse, or a shop after closing time. However, search strategies of this type can also be used in search and rescue missions, or when looking for an item that might be moved by a non-adversarial agent in a larger area, such as a warehouse.

This problem was first proposed by Suzuki and Yamashita and later studied by e.g. Gerkey, Hollinger, and Kolling. Guibas et.al. showed that the problem is NP-hard, a fact that essentially removes the hope of finding optimal solutions in reasonable time.

Good approaches to this problem, in which many pursuers are allowed, seem to be lacking in the literature, especially when the regions are nontrivial. We have developed a new framework, in which the problem is solved as a Mixed Integer Linear Program. As a second step we would like to investigate how heuristic methods can be used within this framework, such as tabu search, simulated annealing, or ant colony optimization, to name a few.

The student should have knowledge of scientific computing. Some familiarity with programming languages such as C or C++ is also appropriate, however not necessary.

Project 2: Autonomous Helicopter Landing on a Mobile Platform

The aim of this Bachelor's Thesis project is to study problems related to autonomous helicopter landings, in particular the problem of attitude control. An interesting scenario is that of landing on a mobile platform, whose motion may or may not be purely rotational, purely translational, accelerating/decelerating etcetera. Examples of situations where such capabilities are required include autonomous landing on the deck of an aircraft carrier in rough seas or on top of a moving vehicle. These two examples are somewhat different: in the former case the platform would mainly be tilting while in the later it would primarily be turning. In both cases, the autopilot should strive to achieve a desired orientation of the helicopter landing gear as it approaches the ground.

The control of orientation, i.e. attitude control, is a research area of current interest at the Division for Optimization and Systems Theory. A new control algorithm allowing a number of mobile manipulators to cooperatively align the attitude of a rigid object with a reference value was developed recently by us. The algorithm was designed to align the object with a static reference attitude but it would be interesting to see how well it fares against a reference attitude that evolves in time and/or look at alternative approaches. That problem could be studied in its own right, but is more interesting in the context of an application, e.g. autonomous helicopter landing.

The student should have knowledge of scientific programming, preferably MatLab, and of classical mechanics corresponding to the basic courses at the engineering physics program. Familiarity with control theory and/or the theory of dynamical systems would also be very useful.