Search

Light detection and ranging (LiDAR) sensors can create high-quality scans of an environment. However, LiDAR point clouds are affected by harsh weather conditions since airborne particles are easily detected. In literature, conventional filtering and artificial intelligence (AI) filtering methods have been used to detect, and remove, airborne particles. In this paper, a convolutional neural network (CNN) model was used to classify airborne dust particles through a voxel-based approach. The CNN model was compared to several conventional filtering methods, where the results show that the CNN filter can achieve up to 5.39% F1 score improvement when compared to the best conventional filter. All the filtering methods were tested in dynamic environments where the sensor was attached to a mobile platform, the environment had several moving obstacles, and there were multiple dust cloud sources.

The division of a city into several waste collection areas can have a large influence on the workload distribution put in place for the waste collection personnel. Additionally, areas with rapid urban development can benefit from a systematic way to automate the collection area division given that there is geographical information containing stop locations with an approximate number of dwelling units. This paper proposes a 2-stage collection area optimization process using the weighted K-means algorithm paired with differential evolution to minimize the standard deviation of dwelling units across each collection area. Results from a case study in The City of Oshawa, Canada prove that the proposed clustering techniques can yield a set of collection areas with 87.75% improvement compared to the current arrangement in terms of balancing the dwelling units. Additionally, the same clustering techniques can be used to assign collection routes for the vehicles in each area. A combination of Dijkstra’s and Hierholzer’s algorithms is applied to generate a route simulation with accompanying statistics regarding the total distance travelled, collection time, travel time, and fuel consumed. Specific to the case study in The City of Oshawa, each day of the week has 2 collection areas, and a genetic algorithm is used to find the optimal collection area pairs. Results from the collection area pairing show that there is a 38.04% and 37.54% improvement of simulated statistics for Week 1 and 2 collections, respectively.

This paper presents an adaptive and weighted model predictive control (MPC) algorithm for autonomous driving with disturbance estimation and prediction. Unexpected and unpredictable disturbances in the real world limit the performance of MPC. To overcome this limitation, this paper proposes adaptive and weighted prediction methods with a sliding mode observer and a weighting function with the grey prediction model. The sliding mode observer is designed for disturbance estimation with finite stability conditions, and the estimated disturbance is predicted using the grey prediction model. Based on the adaptive and weighted prediction method, the length of prediction horizon and cost value of each predicted state are adjusted in real time to eliminate any negative impact on future predicted states. Meanwhile, a variation in the cost value, which is caused by prediction horizon adaptation and weighted prediction, may harm the control performance as it can excessively increase or decrease the model uncertainty. Therefore, an input weighting factor is adapted in the MPC cost function based on an exponential weighting function. The performance of the proposed adaptive control algorithm is evaluated using CarMaker software under longitudinal and lateral autonomous driving scenarios.