The Korean ITS solutions company Nota has launched a solution focussing on intersections, calling them the core of urban problems, and proposing a management system to make them more efficient and safer through edge AI techniques.
It says to build faster and safer intersections, Nota is offering Artificial Intelligence traffic signal control and safe crossing solutions.
“AI Camera is a vision analysis technique to collect traffic data from CCTV in the intersection,” it explains. “It can collect various traffic data (such as traffic flow, time-mean speed, occupancy, etc) similar to the existing, passive sensors such as loop detectors and radar. However, it is easier to install and more cost-effective for maintenance compared to the current sensors.”
Nota says the benefits of its camera solution are summarised as Edge AI efficiency. It can process multi-channel video information in real-time, even within the on-site (low performance) edge board supported by Nota’s AI lightweight technology, which it calls on-site and real-time feasibility. “Due to our edge AI technology, AI Camera can reduce the cost of data communication and latency compared to the other server-based vision analysis methods,” it adds. “Second, Nota AI Camera is powerful because it can classify vehicles into seven types and secure over 95% accuracy in various weather conditions (daytime, night, sunrise, sunset, snow. Rain, etc). In addition, it also collects spatial information (such as space-mean speed) of the road using the perspective transform to improve the data applicability to the various fields.”
AI Traffic Signal Control: “Multi-agent cooperative signal control solution through Multi-agent Reinforcement Learning”
Furthermore, the company says AI Traffic Signal Control is an AI-based signal control method based on real-time traffic data. The benefits of AI Traffic Signal Control are that is possible to respond immediately to the dynamic changes of the roadways, and provides optimal signal control solution in real-time through a data-driven method.
“At first, AI Traffic Signal Control uses traffic data collected from AI Camera as its input variable,” it adds. “Traffic data, such as the queue length, inflow, outflow, speed, is considered as the traffic state, which is the input variable of the algorithm. Next, AI Traffic Signal Control yields optimal signal control solution in real-time through reinforcement learning (RL). It gives the multi-stage optimal solution in real-time (within one signal cycle) responding to the current situation of the intersection. In addition, it also provides the cooperative solution for the traffic network with multiple intersections through multi-agent reinforcement learning (MARL) by rebalancing the traffic flows on the network. AI Traffic Signal Control can provide an optimal solution for an isolated intersection and for arterial roadways and traffic networks.”
AI Safe Crossing: “Predict and prevent the risk of accidents”
AI Safe Crossing is an intersection safety system that can prevent accidents by alarming the hazardous situation in the intersection. This system consists of three functions, Real-time object tracking, Incident prediction, and Alarming. At first, it detects the objects (vehicles, pedestrians, etc.) and tracks the trajectory of each object in real-time using edge AI vision analysis of AI Camera. Next, it predicts hazardous situations through the short-term trajectory prediction. It predicts each object’s 1~3 seconds future trajectories and determines an incident when two or some trajectories are conflicted. Finally, AI Safe Crossing gives alarm for hazardous situations combined with some safety guidance devices, such as LED lights and speakers.
Nota adds, “Nota AI Safe Crossing can effectively prevent accidents since it provides selective information for hazardous situations, enhancing the system’s reliability before human perception-reaction time (PRT). We expect AI Safe Crossing can prevent accident scenarios such as signal violations, poor visibility due to surrounding obstacles, and even dilemma situations for right-turning vehicles. This system is expected to play an important role in C-ITS intersection’s safety system by expressing less than 1-second short-term information on the LDM (Local Dynamic Map) by processing video data in real-time with Edge AI without V2X I2X communications.”
(Pictures – Nota)