Software-defined vehicle (SDV) architectures are fundamentally changing how electric vehicles (EVs) allocate compute, manage energy, and coordinate increasingly interdependent systems.
Severin Bredahl-Banovic, VP, Head of Business and Solution Architecture, HERE Technologies
As centralized computing platforms take on propulsion, battery management, ADAS, and perception workloads simultaneously, EV-specific power and thermal constraints are becoming just as critical to system design as raw performance.
Severin Bredahl-Banovic, VP, Head of Business and Solution Architecture at HERE Technologies, brings a system-level perspective on how location intelligence is applied within SDV architectures. HERE provides digital mapping and location data used across automotive, mobility, and logistics applications.
In this Q&A, Bredahl-Banovic discusses how EV power and thermal constraints shape SDV compute strategies, how centralized architectures coordinate battery, propulsion, and ADAS domains, and why predictive, map-based context is increasingly used to manage perception load, redundancy, and functional safety in modern EV platforms.
Here’s what he has to say…
As EVs adopt software-defined vehicle (SDV) architectures, what are the main challenges in coordinating propulsion, battery, and ADAS systems under a centralized compute platform?
Severin Bredahl-Banovic: Centralized compute is the cornerstone of SDV, but aligning propulsion, battery, and ADAS systems under one architecture is complex. A rich location intelligence layer addresses this complexity by delivering high-quality maps full of attributes such as lane topology, traffic rules, and elevation.
These attributes enable precise orchestration across domains, ensuring energy management and automated driving decisions are context-aware and synchronized for safety and efficiency.
Centralized SDV architectures rely on shared data layers to coordinate propulsion, battery management, ADAS, and vehicle software, with location intelligence providing predictive context across control domains.
How do EV power and thermal limits shape the compute capabilities needed for SDV features?
Bredahl-Banovic: EVs must balance compute power and propulsion for energy use. To mitigate this, OEMs must shift intelligence from hardware to software and the cloud. Intelligent, software-native routing algorithms leverage rich map attributes to help an OEM dynamically prioritize ADAS features based on route complexity and energy constraints.
Energy consumption can only be accurately determined when projected over a specific route. By integrating with energy consumption models, these systems can predict and optimize routes using topography details such as curvature, elevation, and slope.
How should SDV compute systems resolve inconsistencies between sensor data and map information?
Bredahl-Banovic: Sensor discrepancies are inevitable in dynamic environments and resolving them is critical for safety and performance. To address this challenge, sensors and maps must work in concert, each complementing the other. Sensors provide real-time situational awareness, while high-quality, attribute-rich maps deliver predictive context that sensors alone cannot anticipate, such as road geometry, traffic rules, and surface conditions beyond the immediate line of sight.
Vehicles often face discrepancies between sensor readings, map data, and road conditions. To address this, systems need redundancy and strong feedback loops. Maps serve as a collective memory, correcting sensor errors and updating with new sensor information to improve accuracy. This ongoing process helps SDVs make safer, more reliable decisions in uncertain environments.
How can SDV platforms manage increasing perception and sensor-fusion data loads while maintaining reliable real-time performance?
Bredahl-Banovic: While sensor fusion powers autonomous systems, onboard sensors have limitations in their spatial coverage and range of vision. Rich attribute maps act as a predictive layer, expanding the vehicle’s range of vision while reducing reliance on raw sensor processing. Maps provide pre-processed, localized spatial information.
Energy-aware routing in software-defined EVs combines real-time vehicle data with predictive location intelligence, such as charger availability and traffic conditions, to manage range, compute load, and ADAS behavior.
By relying on this data for features that don’t change in real time, SDVs can reserve more computing power for processing dynamic events. By supplying contextual cues such as road curvature, road signage, and surface conditions, location intelligence helps OEMs optimize perception workflows and maintain real-time responsiveness.
What redundancy strategies help keep critical functions stable when compute resources become constrained?
Bredahl-Banovic: Stable operational design is essential for SDVs. Redundancy is strengthened through multi-source map updates that are pre-processed, saving resources in the vehicle and ensuring navigation and ADAS continuity even under compute constraints.
These solutions enable graceful degradation, preserving steering, braking, and situational awareness when systems are stressed. In SDVs, resilience is the new reliability.
How should software handle degraded perception or connectivity inputs while preserving functional safety in EVs?
Bredahl-Banovic: Effective software anticipates these scenarios with offline map capabilities and predictive fallback maneuvers. Location intelligence can anticipate upcoming connectivity gaps, allowing vehicle systems to proactively cache information before entering areas with limited coverage.
It can also predict and localize conditions that degrade perception, such as fog or heavy rain, so drivers can take control from ADAS functions to adjust their speed and driving behavior accordingly.
What issues emerge when multiple control domains depend on a shared middleware layer in EVs?
Bredahl-Banovic: Shared middleware simplifies integration, but introduces risks such as resource contention, version conflicts, and potential security gaps. These are best mitigated through containerized services and modular SDKs, isolating map and sensor data layers for stability and security. This ensures that as control domains converge, they do so without compromise.
Filed Under: Featured Contributions, Q&As, Software