Mobility AI

May 12, 2023June 21, 2025 byMWB Team

Edge AI with IoT in Smart Mobility Solutions

The convergence of Edge AI and the Internet of Things (IoT) is revolutionizing smart mobility solutions by enabling real-time, low-latency decision-making critical for Intelligent Transportation Systems (ITS). This synergy is essential for applications like vehicle-to-everything (V2X) communication, autonomous vehicles, and smart parking systems, where responsiveness and scalability are paramount.

Before we dive deep:

Edge AI: Edge AI refers to deploying artificial intelligence (AI) algorithms directly on edge devices, such as sensors, cameras, and embedded systems, rather than relying on cloud computing. Key benefits include reduced latency, enhanced privacy, and real-time processing.

Internet of Things (IoT): IoT connects physical devices, enabling them to collect and exchange data. When integrated with Edge AI, IoT systems can process data locally, making them more efficient and responsive.

Why Edge AI for Smart Mobility?

Smart mobility solutions demand immediate responses to dynamic conditions, such as traffic congestion, pedestrian crossings, and emergency vehicle prioritization. Edge AI enables low latency in processing data at the source minimizing delays, high reliability wherein the systems can operate independently of network connectivity, and scalability as the local processing reduces the load on centralized cloud resources.

Applications of Edge AI and IoT in Smart Mobility

Vehicle-to-Everything (V2X) Communication: V2X encompasses communication between vehicles, infrastructure, pedestrians, and the network. Edge AI enhances V2X through Collision Avoidance by processing the sensor data in edge devices installed in the vehicles, Traffic Signal Optimization with AI adjusted timings dynamically based on the real-time traffic conditions and Hardware Platforms e.g. NVIDIA Jetson AGX Xavier is commonly used in V2X systems for real-time perception and decision-making.

Autonomous Vehicles: Autonomous vehicles rely heavily on Edge AI to process vast amounts of data from LiDAR, cameras, and other sensors. The outcome is used for Object Detection and Tracking wherein Edge AI processes camera feeds to identify pedestrians, vehicles, and obstacles, Path Planning to find optimal routes based on the road conditions as per the real-time algorithms calculations and Hardware Platforms which includes devices like NVIDIA Jetson Orin and Intel Movidius to facilitate high-performance AI computation at the edge.

Smart Parking Systems: Edge AI transforms parking infrastructure by enabling Occupancy Detection wherein IoT sensors embedded in parking spots detect availability in real-time, Dynamic Pricing to optimize pricing based on demand and occupancy rates and Hardware Platforms like Raspberry Pi 4 with AI accelerators like Coral TPU provides cost-effective solutions for parking management.

Technical Architecture

A typical Edge AI-IoT architecture for smart mobility includes:

Sensors: LiDAR, cameras, and ultrasonic sensors collect raw data
Edge Devices: Devices like NVIDIA Jetson process data locally, running AI inference models.
Connectivity Modules: 5G, DSRC, and LoRa facilitate communication between devices and infrastructure.
Cloud Integration: While primary processing occurs at the edge, the cloud is used for long-term storage and analytics.

However the real challenge lies in Hardware Limitations because of the limited computational power in the Edge devices compared to the cloud systems, which can be addressed by specialized hardware accelators like GPUs and TPUs designed for AI workloads.

Another challenge lies in the data privacy and security because the Edge AI systems must secure the sensitive data locally. In addition, integrating thousands of devices in urban environments can strain resources. However, with the advent of newer encryption protocols the boot processes can be secured on edge devices and lightweight AI models can be leveraged for scalability.

“At MWB, we see the convergence of Edge AI and IoT as a pivotal force in shaping smart urban mobility—creating systems that not only adapt in real-time but also set new benchmarks for efficiency, safety, and sustainability.”
Director, ITS & TECH – MWB

Future Trends in Edge AI and IoT for Smart Mobility

We, at MWB, believe that the learning in this space is going to evolve tremendously in the coming years. We see already “Federated Learning”, which enables collaborative training of AI models across edge devices without sharing raw data. There is enormous amount of research and development in the area of Energy-Efficient AI for the development of low-power AI models to extend the battery life of edge devices. The “Integrated V2X Ecosystems” allow Seamless integration of edge AI and IoT with urban digital twins for predictive traffic and infrastructure management.

For further discussions around this and to know more about it, reach out to us at info@mwb-me.com

Mobility AI

March 7, 2023January 13, 2025 byMWB Team

Predictive Maintenance with AI For Urban Mobility Systems

Urban mobility systems are the lifelines of modern cities, encompassing infrastructure such as roads, bridges, tunnels, and transit networks. Maintaining these systems efficiently is paramount to ensuring uninterrupted operations, safety, and cost-effectiveness. Predictive maintenance, powered by Artificial Intelligence (AI), has emerged as a game-changer, enabling infrastructure managers to anticipate failures and optimize maintenance schedules.

Predictive maintenance involves the use of advanced analytics, machine learning (ML), and IoT sensors to monitor the condition of assets in real-time. By analyzing historical and real-time data, predictive models identify patterns and anomalies that signal potential failures. This approach contrasts with traditional reactive maintenance, which addresses issues after they occur, and preventive maintenance, which relies on fixed schedules irrespective of actual asset conditions.

Core Components of AI-Driven Predictive Maintenance

IoT Sensors and Data Collection:
Sensors capture diverse data streams such as temperature, vibration, pressure, and structural stress in real-time. For example, accelerometers and strain gauges embedded in bridges detect vibrations that could indicate structural weaknesses.
Machine Learning Algorithms: ML models process and analyze data to identify degradation patterns and predict time-to-failure. For example, Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks are used for time-series analysis in dynamic systems like escalators or traffic signals.
Digital Twins: Digital twins—virtual replicas of physical assets—simulate operational conditions to forecast failures and test maintenance strategies. For example: A digital twin of an urban rail system can model wear and tear based on real usage patterns.
Centralized Monitoring Platforms: SCADA and cloud-based platforms aggregate and visualize data, allowing operators to monitor infrastructure health remotely. Integration with AI enhances decision-making by providing actionable insights.

Applications in Urban Mobility Systems

AI-powered sensors monitor structural integrity, detecting cracks or corrosion before they escalate in the wide spectrum of urban infrastructure, including tunnel and bridges, traffic signal systems, public transit and roadway infrastructure. Predictive models optimize inspection schedules, reducing costly manual checks, downtime. Anomaly detection algorithms predict hardware failures, enabling timely replacements. Pavement sensors and ML algorithms can also predict pothole formation based on traffic load and environmental conditions for proactive repair schedules.

A Technical Glance at AI in Predictive Maintenance

The anomaly detection is usually carried out by identifying the outliers in sensor data using techniques like k-means clustering or autoencoders. For predictive modelling, failure probabilities can be forecasted through regression models, decision trees, and neural networks. The maintenance schedules can be further optimized by reinforcement learning from simulated environments.

However, the high frequency sensor data requires scalable storage solutions, which are often cloud based. The data quality remains as one of the biggest challenges which needs to be addressed by preprocessing techniques such as imputation and smoothing. The learning from the predictive models must integrate seamlessly with legacy systems and SCADA platforms, requiring robust APIs and middleware, which must be planned in advance.

“AI in predictive maintenance isn’t just about anticipating failures – it’s about redefining the way we maintain, operation, and extend the life of critical infrastructure.”
Head of ITS & TECH, MWB

What we can expect in future

Based on the findings from the research and prototype of predictive maintenance project through a strategic collaboration of MWB, we can expect significant accuracy (>80%) in predicting motor failures to reduce downtime (<20%) in addition to a estimated lower operational costs by 15% within the first year itself.

With the future technologies around Edge Computing (real-time analytics at the edge itself to reduce latency and improve responsiveness), Hybrid AI Models (combined physics-based models with ML for greater predictive accuracy) and Sustainability practices (reduced resource waste and energy consumption), we can only expect that AI driven predictive maintenance will only get better, more accurate and the DNA of urban infrastructure technology.