Journal

E-readers are a popular way to read books, but some people may find the computer-generated voices used to read the books to be unnatural or robotic. Using trained neural networks to generate more human-sounding voices could certainly improve the reading experience for many people.

The ROBOCART project developed a fully-automatic simulated supermarket cart, named ROBOCART (tracker), to provide customers with a hands-free shopping experience. The system features human-tracking, trajectory prediction, automatic path-planning, and automatic charging capabilities. The project utilized an A* algorithm to plan the cart’s path and compared it with the most optimal path, explaining their design choices. The Kalman-Filter was also integrated into the system to make the customer-following task smarter and demonstrated how it saves up to 13% of the time in certain scenarios.

In the field of robotics and artificial intelligence, the primary task is to determine the mobile robot’s location, known as ‘Localization.’ Several methods can be used for mobile robot localization, such as the Kalman filter, Grid-based Markov Localization, and the Monte Carlo Localization (Particle filter). The Monte Carlo method or Particle filtering is currently the most popular localization method due to its optimal trade-off between accuracy and robustness. This method involves describing a probabilistic distribution through a sampling process. The map is sampled based on known prior probabilities, and each particle is assigned a weight representing its probability. Through iterations, the estimate pose of the robot accurately converges.