Home > Center for Intelligent Mobility (CIM) > Our Projects

Our Projects

  • CIM

Advancing Intelligent Transport

The Center for Intelligent Mobility (CIM) at KLE Tech is a pioneering hub for innovation in autonomous systems, intelligent connectivity, and vehicular cyber-physical security. At the heart of our mission is the creation of safer, smarter, and more sustainable mobility ecosystems.

CIM integrates AI-driven perception and control, secure communication protocols, and rigorous real-world testing to solve next-generation mobility challenges. Whether it’s navigating complex urban traffic, enhancing vehicle-to-vehicle and vehicle-to-infrastructure communication, or ensuring system-level security, our work is grounded in practical impact and scalable design.

Through strategic collaborations with industry leaders like NVIDIA, L&T Technology Services (LTTS), and BOSCH, we ensure that research outcomes go beyond the lab—fueling real-world innovations that shape the future of intelligent transportation.

cim-advance-intelligence-banner
  • RESEARCH PROJECTS AT CIM

Vision in Action

At the Center for Intelligent Mobility (CIM), our projects are where cutting-edge research meets real-world application. Built on a foundation of AI, deep learning, and sensor fusion, our work tackles pressing mobility challenges—from enhancing autonomous navigation to enabling high-precision spatial perception.

Our approach is hands-on and collaborative. Every project blends experimental validation with system-level thinking, constantly pushing the boundaries of connected, autonomous, and intelligent transportation technologies.

  • BEVFlow – Streamlined Excellence in Instance Prediction from a Bird’s-Eye View
  • BEVFlow addresses a key challenge in autonomous driving—accurate motion prediction—using a novel approach that fuses centripetal motion modeling with backward flow warping. It processes inputs from multi-view cameras to generate Bird’s-Eye View (BEV) feature maps, which are used to predict future motion flows and segmentation maps.
    A dedicated warping-based post-processing module enables high-precision instance reconstruction. The project focuses on both accuracy and efficiency, incorporating iterative improvements through hyperparameter tuning and training refinement. BEVFlow is a step forward in enabling safer, more context-aware autonomous systems.
  • Affordance Learning for Detection of Drivable Area
  • This project focuses on identifying drivable regions using affordance-based deep learning. Leveraging data from LiDAR, radar, and camera sensors, neural networks are trained to infer high-level driving cues—such as lane boundaries, obstacle distances, and free space—from raw sensory input.
    By fusing spatial and temporal features via advanced convolutional and recurrent architectures, the system enhances situational awareness and real-time decision-making. Robustness and generalization across environments are core to the development, making it highly applicable to complex and dynamic driving scenarios.
  • ReConfNET – Confidence-Driven SLAM for Scene Reconstruction
  • ReConfNET is an RGB-D SLAM framework designed for high-fidelity indoor scene reconstruction. It combines a hierarchical generative model with a confidence-based loss function to improve both pose estimation and depth prediction.
    The system processes continuous RGB-D input, selectively extracting high-confidence key points. It utilizes neural implicit encoding and volume rendering to produce scalable and accurate 3D reconstructions. ReConfNET outperforms traditional SLAM systems in both accuracy and computational efficiency, opening up new possibilities for robotics, augmented reality, and intelligent spatial environments.
  • RESEARCH FOCUS AREAS

Domains of Impact

01

Perception and Control

02

Security

03

Connectivity