Lars Hammarstrand

Title: Lars Hammarstrand: Inventor in Vehicle Road-Geometry Estimation.

Introduction.

Lars Hammarstrand is an inventor based in Mölnlycke, Sweden. His patent work focuses on estimating road geometry for vehicles, including methods that use machine learning and vehicle sensor data. He is associated with Zenseact AB and has 2 patents.

Latest Patents.

Road Geometry Estimation for Vehicles. This invention describes a method for representing road objects around a moving vehicle. It processes camera images across consecutive time steps, uses machine-learning algorithms to identify image features, and transforms those features into a Bird’s Eye View representation. Transformer-based algorithms then extract object embeddings, which are decoded to determine each road object’s position and class.

Model-Based Road Estimation. This invention presents a computer-implemented method for estimating the geometry of a road on which a vehicle is traveling. Its model-based approach can represent roads with varying numbers of lanes, as well as entrance and exit lanes. It uses Bayesian multi-lane tracking and combines vehicle-motion data with perception data from sources such as cameras, lidar, and radar to estimate lane-center curves and account for detected road features and other road users.

Career Highlights.

Hammarstrand’s listed patent work addresses a key challenge in vehicle technology: building reliable representations of complex road layouts. The inventions describe complementary approaches, combining image-based object detection with model-based estimation of multi-lane road geometry.

Collaborations.

Hammarstrand is associated with Zenseact AB. Junsheng Fu and Ossian Eriksson are listed as coworkers; the available information does not specify their roles in these patents.

Conclusion.

Lars Hammarstrand’s patent portfolio centers on vehicle road-geometry estimation, with approaches that use camera imagery, sensor data, and machine-learning techniques to represent road objects and lane structures.