Rana Alkhoury Maroun
Title: Using the Mushroom Body Architecture for Locality-Sensitive Hashing
Abstract: Locality-sensitive hashing (LSH) is a technique used to produce hashes that preserve the distances between the inputs. This allows to bin similar inputs together, making it much faster to search for approximate nearest neighbors in large datasets. Recently, models inspired by the Mushroom Body (MB) outperformed previous LSH functions at preserving Euclidean distances while being less computationally expensive. In this presentation, I will talk about how different MB-inspired architectures affect the accuracy of the LSH function, and explore the possibility of preserving perceptual (instead of Euclidean) distances, which could be more relevant to real world applications.
In this workshop I am going to talk about some of our recent work on designing an efficient multi-agent cooperation technique for different levels of applications.
For Search and Rescue (SAR) applications, we considered the use of a team of multiple unmanned aerial vehicles (UAVs) to accomplish the mission where the SAR mission should be planned in a way that focuses more attention on the centre, where most of the survivors are located. The simulation results show that the UAVs running the Layered Search and Rescue (LSAR) algorithm rescue approximately 8% higher and 77% faster than the next best algorithm. To conclude, the main factor to be optimized in the SAR problem concerns how early the rescue mission commences.
For Mobility on Demand (MoD) applications, we consider the use of a team of multiple unmanned ground vehicles (UGVs) to meet the transportation demands in future cities. To demonstrate the feasibility of the approach, an Autonomous Mobility on Demand (AMoD) model is analyzed in different cities in comparison to the traditional transportation systems. The results for ten case studies from the literature confirmed that a shared AMoD fleet has the potential to reduce the number of vehicles necessary to deliver the same traveling demand by 85%. Moreover, analytical results illustrated that AMoD systems will remove the need for on-street parking spaces drastically. On the other hand, the results also showed that there are likely to be some negative impacts such as increased total kilometers of travel due to repositioning and fleet rebalancing tasks. The realistic model of the city of Edinburgh would be analysed in the same manner for future work.