ALUMNI

TOPIC : OPTIMAL PATH PLANNING FOR ROBOT NAVIGATION

          Path planning is major field of robot navigation and has been a key area of research over the years . Among them sampling based algorithms have been in spotlight because of their better computational cost and fast convergence but they do not guarantee optimal path . Graph based algorithms like Visibility graph assure better path planning but have large computational cost specially in environments with large number of obstacles .. Therefore we have proposed LTA*, a new technique to plan optimal path with less computational cost . It uses concept of Local tangents to optimally plan path especially in environments with concave regions in less time as compared to other approaches . We have also proposed a simple corner selection technique that selects only convex corners which reduces computational cost of Algorithm . As LTA* is based on A* algorithm therefore path planned from LTA* assures that it is shortest (Optimal) . Results have proven that performance of LTA* is independent of map size and obstacles type.
Supervisor : Dr . Wajahat Hussain .

TOPIC : PATH PLANNING IN UNKNOWN ENVIRONMENTS USING LINGUISTIC INSTRUCTIONS AND PICTORIAL CUES

          Verbal navigational instructions have shown to be helpful in exploring unknown areas . In this work, we propose the use of pictorial cues (sketches) to aid navigation in unknown areas . These sketches can be used to differentiate between 1 ) patterns which are difficult to describe and easier to draw 2 ) objects with fine grained differences and 3 ) novel objects with no training data and common knowledge . We propose a novel “draw in 2 D and match in 3 D” algorithm to make sketch matching viewpoint invariant . We show superior performance of our sketch aided navigation on standard datasets . We also provide a challenging Crossroads dataset . This dataset contains both indoor and outdoor scenes with fine grained differences between scenes
Supervisor : Dr . Wajahat Hussain .

TOPIC : ROBUST PLACE RECOGNITION IN EXTENSIVELY CHANGING ENVIRONMENT FOR ROBOT NAVIGATION

          Visual place recognition using hand crafted and deep features performs well in static environments . The dynamic environments with extensive changes which are very common are however difficult to be recognised . The environments may vary in appearance due to many reasons : weather changes, seasonal changes and changes in lightning conditions Visual place recognition can be incredibly enhanced if it becomes possible to estimate the appearance of a specific scene at a specific time in view of the appearance of the scene earlier and learning the way in which appearance vary over time . In this thesis, we examined whether worldwide appearance changes in an environment can be learned adequately to enhance place recognition . We used day night pairs for training a learned model using cGANs that efficiently approximates a night scene based on a day scene . We have used binary descriptor based on color histograms for image matching . The experiments have been done on three datasets collected from different environments . The experimental results show that the visual place recognition with images approximated by the trained model outperforms the visual place recognition based on raw images and currently available state of the art methods
Supervisor : Dr . Wajahat Hussain .