No description available.

Autonomous lawn mowers and floor cleaning robots are today easily accessible and areutilizing well-studied Coverage Path Planning algorithms. They operate in single-floorenvironments that are small with simple geometry compared to general urban environments such as city parking garages, highway bridges or city crossings. A next step for autonomous cleaning is road sweeping of these complex urban environments. In this work,a new Coverage Path Planning approach, Sampled BA* & Inward Spiral , handling this taskwas compared with existing well-performing algorithms BA* and Inward Spiral. The proposed approach combines the strengths of existing algorithms and demonstrates state-of-the-art performance on three large-scale 3D environments. It generated paths with lessrotation, while keeping the length of the path on the same level. For a given starting point,the new approach had consistently lower cost (length + rotation) for all environments. Forrandom starting points, randomness in the new approach caused less robustness, givingsignificantly higher cost. To improve the performance of the algorithms and remove biasfrom manual tuning, the parameters were automatically tuned using Bayesian Optimization. This makes the evaluation more robust and the results stronger.

The use of micro aerial vehicles (MAV) for civilian use such as exploration and inspection of varying structures, equipment and areas have garnered some interest as of late. MAVs have the mobility and agility to traverse three dimensional space quickly and access hard to reach areas where other alternatives would struggle, but a flying platform such as a MAV comes with it’s own set of distinct problems. Almost any collision with the environment results in a complete failure of the platform. Any exploratory framework would need to perform obstacle avoidance and online path planning in a fully unknown environment with low computation times to ensure that the limited battery resources on the MAV is used in the most efficient way. In this thesis the exploratory rapidly-exploring random tree (ERRT) framework will be further optimized and an efficient strategy for finding valid exploration paths which are not in the immediate vicinity of the MAV will be developed and integrated. The method is demonstrated and proven through results from simulations and real life experiments.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.