Diagnostics and Explainable Machine Learning Models

Despite the advances of Machine Learning, the models are still being considered black-boxes that are difficult to diagnose and explain. The model performance diagnostic measures are critical to the assessment of the model’s relevance, accuracy and robustness. Good models’ performance is the primary enabler of their successful deployment in real-life applications. However, even if the models perform well, it is not known why the models predict the way they do, that is, which input variables are responsible for the models’ predictions.

Deployment of motion platform control architecture for a high-fidelity driving experience in simulators

Realistic driving experience in motion simulators is a key element for the impressiveness of the VR based simulators. In the driving simulator the free motion of the vehicle is mapped to a motion platform with limited workspace by filtering out the motion and scaling it down with a motion cueing algorithm. These algorithms should be designed in such a way to give a feeling to the users as if they are driving a real vehicle.

Integrated Optimal CLT Building Design Considering Energy Performance and Structural Performance

This research proposed a method to conduct optimal CLT building design, which has minimum CLT usage and minimum operating energy consumption. This is a multidisciplinary research pushing the boundary of current definition of optimal design for each disciplinary to another level. With the developing of Building information modeling and automized simulation-optimization technology, the design of building will evolve to an interdisciplinary design in this decade, and this research demonstrate the benefit and application of integrated design.

Towards an Elastic and Reliable Cloud Resource Management

This work is a holistic automatic methodology for cloud resource management system, which is a corner stone to build any cloud system. Cloud players rely on this to reduce management effort and cloud running cost, by enabling dynamic service access to cloud clients with cheapest price for customer, and high revenue for cloud providers. Customer requirements must be achieved based on their service level agreement like availability of the service.

Pickup and Delivery with Smart Service Times

The research problem to be addressed is the pickup and delivery problem (last mile) with precedence constraints for the pickup/delivery operations, and time windows. The objective is: (i) the design of a new algorithm that will use the algorithm of Curtois et al.

Development and Validation of a Protein-Protein Interactions Modeling Platform for Rapid Affinity Predictions and Pharmaceutical Applications

Developing a drug for new diseases cannot only be challenging but also time consuming. From the identification of a druggable target to a compound which can improve a condition it usually takes more than 12 years. Since there is basically an infinite number of possible compounds which can be turned into a drug it is literally the problem of finding a needle in a haystack. The trial and error method of making molecules in the laboratory and testing their efficiency has been proven successful for over a century.

Machine Learning for Modeling Soil-Tool Interactions

When construction equipment digs, this generates a complex set of physical interactions between the machine and the soil. Being able to accurately simulate such interactions in real-time opens the door to improved operator training and even adaptively-tuned digging operation that optimize the energy-efficiency of construction equipment. With recent advanced in artificial intelligence (AI), this is now within reach.

Automatic matching of service offers to calls for tenders

The process for responding to tenders is very long, arduous and involves many repetitive tasks. Among these, it is necessary to read the invitation to tender in order to find the different requirements, important dates, legal obligations, constraints of presentation of the response, etc. All of these factors are used for the prequalification of a tender and for the drafting of the tender. These requirements are generally expressed in the form of free text.

Novelty Detection in Lunar Analogue Terrains

Lunar and planetary rovers are faced with a high volume of data from their sensors and cameras, and yet decisions must be made rapidly to prioritize to which crater it should drive up to and of which rock it should take a closer look. In Canada’s intended upcoming exploration of the Lunar surface aboard short-timescale commercial missions, these considerations are of utmost importance.

Real-time surrogate safety analysis solution using Lidar technology powered by artificial intelligence

The safety of intersections, interchanges, and other traffic facilities is most often assessed by tracking and analyzing police-reported motor vehicle crashes over time. Given the infrequent and random nature of crashes, this process is slow to reveal the need for remediation of either the roadway design or the flow-control strategy. This process is also not applicable to assess the safety of roadway designs that have yet to be built or flow-control strategies that have yet to be applied in the field.

Pages