Autonomous Vehicle Sensor Fusion Where Satcom Positioning Fits Alongside Lidar and Camera Systems
Autonomous Vehicle Sensor Fusion Where Satcom Positioning Fits Alongside Lidar and Camera Systems
Satellite positioning is not the sensor that keeps an autonomous vehicle from hitting a pedestrian. That job belongs to lidar, camera and radar working in the range of milliseconds. Satcom and GNSS positioning play a different but equally critical role: they provide the vehicle's global, absolute location and a timing reference that every other sensor's data gets stamped against. Sensor fusion architectures that treat satellite positioning as an occasional convenience, rather than a continuously managed input with its own failure modes, tend to be the ones that break down in tunnels, dense cities, or contested electromagnetic environments. StarWin builds the communication, navigation, remote sensing and compute infrastructure underneath autonomous platforms, including GNSS anti-jamming antennas and multi-orbit satellite terminals now being evaluated inside vehicle sensor stacks through its work on next-generation autonomous platform integration. That gives the company a working view of where satellite positioning genuinely helps a fusion stack, and where engineers should not expect it to.
TL;DR
· Lidar and camera handle immediate, relative obstacle detection; satellite positioning provides absolute global coordinates and timing that anchor the whole fusion stack.
· Standard GNSS update rates of 1 to 10 Hz create latencies too slow for direct vehicle control, so IMUs and odometry bridge the gap between satellite fixes.
· Urban canyons, tunnels and jamming all degrade GNSS differently, which is why redundancy and anti-jamming design matter more than raw accuracy specs.
· ISO 26262 and ISO 21448, plus regulations like GB 44721-2026 and UN Regulation No. 157, require multi-sensor redundancy rather than reliance on any single positioning source.
· Anti-jamming and anti-spoofing built into the antenna, not added as an accessory, determine whether satellite positioning stays trustworthy when it matters most.
About the Author: This article is published by StarWin, a Chengdu-headquartered provider of compound communication, navigation, remote sensing and computing systems whose CRPA anti-jamming antennas and multi-orbit terminals support next-generation autonomous platform integration work.
What Does Sensor Fusion Actually Mean for Autonomous Vehicles?
Sensor fusion is the process of combining data from multiple sensor types into a single, coherent model of the vehicle's position and surroundings, because no single sensor is reliable enough on its own. Lidar measures distance to objects using laser return time, cameras capture visual texture and classification data, radar detects velocity and works through fog or rain, and GNSS supplies an absolute position on the globe. Each sensor fails differently: cameras struggle with glare and darkness, lidar degrades in heavy precipitation, and GNSS can lose lock entirely under obstruction. Current production automotive lidars typically deliver ranges of 200 to 300 meters, frame rates of 10 to 20 Hz, and angular resolution around 0.1 degrees, while cameras run at 30 to 60 frames per second with high-resolution color and texture data. Neither gives the vehicle a global reference frame; that is what satellite positioning is for. Fusion algorithms weight each sensor's input based on confidence, so when GNSS confidence drops, say inside a parking structure, the system leans harder on lidar-based localization and inertial dead reckoning instead of discarding the data outright.
Why Can't Lidar and Cameras Replace GNSS Entirely?
Lidar and cameras answer "what is around me right now," while GNSS answers "where am I on the planet," and a fusion stack needs both answers, not a choice between them. This is a mechanism question, not a preference. Lidar and camera systems build a relative map of nearby obstacles by measuring distance and appearance from the vehicle's own frame of reference; that map is precise for the next few hundred meters but has no connection to a fixed global coordinate. Drive the same vehicle through the same intersection twice and the lidar point cloud looks similar each time, but without GNSS there is no way to confirm the vehicle is actually at that intersection versus a visually identical one across town. Autonomous fleets rely on this global anchor for route planning, geofencing, regulatory compliance and handing off between local perception and map-based navigation. Satellite positioning is the sensor that ties the vehicle's local perception bubble to the actual map the rest of the world uses.
How Accurate Is Satellite Positioning for Autonomous Driving, and Where Does It Break Down?
Modern RTK GNSS systems provide centimeter-level accuracy, typically 1 to 2 centimeters, for autonomous vehicles under open skies, which sounds sufficient until the vehicle enters an environment GNSS was never designed for. In urban canyons and other obstructed settings, multipath errors, signal blockage and non-line-of-sight reception can degrade that accuracy to several meters or cause a complete loss of positioning. Standard GNSS receivers also update at only 1 to 10 Hz, introducing latencies of 100 to 1000 milliseconds, which is too slow for direct high-speed vehicle control on its own. In tunnels or dense urban blocks, GNSS signal can drop out completely for the duration of the obstruction. This is why no serious autonomous driving architecture treats GNSS as a standalone control input. IMUs and wheel odometry are used to bridge these gaps, carrying the position estimate forward using acceleration and rotation data until the satellite fix returns. The practical design implication is that satellite positioning needs to be evaluated on its degradation behavior, not just its best-case accuracy figure.
What Do Safety Standards Actually Require From Positioning Redundancy?
Safety regulators do not mandate a specific fusion architecture; they mandate an outcome, and that distinction shapes how engineering teams should approach positioning redundancy. SAE J3016 defines the six levels of driving automation but deliberately stays silent on sensor architecture or hardware redundancy requirements. The actual engineering obligations come from ISO 26262 for functional safety and ISO 21448 for safety of the intended function (SOTIF), both of which require manufacturers to build in sufficient sensor fusion and redundancy to cover hardware faults and performance limits within the vehicle's defined operational domain. Regulatory frameworks including China's GB 44721-2026 and UN Regulation No. 157 go further for Level 3 systems such as Automated Lane Keeping Systems, requiring that the vehicle perform as safely as a human driver and include fail-safe mechanisms. These regulations are technology-neutral, mandating safety outcomes and performance minimums rather than a prescribed sensor architecture, which in practice pushes manufacturers toward multi-sensor redundancy across cameras, lidar and radar plus backup positioning through IMUs and odometry. In practice, this means a positioning subsystem has to prove not just accuracy, but graceful degradation and a documented fallback path when GNSS quality drops.
How Does GNSS Interference and Jamming Change the Fusion Calculation?
Beyond natural signal loss, GNSS is also vulnerable to deliberate or incidental interference, and that risk changes how a fusion stack should treat positioning confidence in real time. A jammed or spoofed GNSS signal does not necessarily disappear; it can report a confident but wrong position, which is more dangerous to a fusion algorithm than an honest outage because the system may trust bad data instead of falling back to other sensors. This is the argument for building anti-jamming and anti-spoofing capability directly into the positioning antenna rather than treating it as an add-on module bolted onto a standard GNSS receiver. StarWin's CRPA-based antennas use highly integrated digital arrays designed to sit inside the terminal or vehicle housing rather than as an external accessory, with anti-spoofing capability on the higher-tier models. The design logic is straightforward: interference detection has to run continuously and feed a confidence score into the fusion algorithm the same way multipath or obstruction does, not act as a separate system that only intervenes after something has already gone wrong.
Where Do Satcom Terminals Fit Beyond Positioning Itself?
Positioning is one input, but autonomous fleets also need connectivity for fleet coordination, remote diagnostics, and data offload, which is where satellite communication does work distinct from GNSS. A vehicle operating outside cellular coverage, on a mine haul road or a remote logistics corridor, still needs to report position, health data and sensor logs back to a fleet operations center. This is a communication problem, not a perception problem, and it is best solved by a terminal that can move between satellite orbits and terrestrial networks depending on what is available. StarWin's multi-orbit approach, reaching GEO, MEO and LEO from a single terminal, combined with automatic roaming between satellite and terrestrial 4G/5G, is built for exactly this kind of fleet-level connectivity rather than moment-to-moment obstacle avoidance. The company's satellite IoT line, running on the TianQi LEO constellation, adds a low-power option for fleets that need periodic position and status reporting rather than continuous broadband, which matters for vehicles where every watt of power budget competes with sensor and compute loads.
How Should Engineers Weigh Positioning Investment Against Lidar and Camera Investment?
The honest answer is that these are not competing budget lines, because they solve different problems on different timescales. Lidar and camera systems govern the vehicle's immediate safety envelope, operating in the range of tens of milliseconds. Satellite positioning governs the vehicle's relationship to the outside world: which road it is on, whether it has left a permitted zone, and what timestamp to attach to every other sensor's data. Underinvesting in positioning integrity does not make the vehicle crash into the car ahead; it makes the vehicle's map-based decisions unreliable, which shows up later as routing errors, geofence violations, or corrupted data logs that undermine incident investigation. A fusion architecture that treats GNSS as a commodity input, rather than a component with jamming resistance, multipath handling, and coordinated multi-orbit or multi-network fallback, is optimizing for the wrong failure mode.
Frequently Asked Questions
Can lidar alone replace GNSS in an autonomous vehicle?
No. Lidar builds a relative map of nearby obstacles from the vehicle's own frame of reference; it has no inherent connection to a global coordinate system, which GNSS provides.
Why does GNSS accuracy vary so much between open sky and urban environments?
Centimeter-level RTK accuracy assumes clear line-of-sight to multiple satellites. Buildings and structures cause multipath reflections and signal blockage, degrading accuracy to several meters or causing total signal loss.
What bridges the gap when GNSS signal drops in a tunnel?
Inertial measurement units and wheel odometry carry the position estimate forward using motion data until the satellite signal returns.
Do safety regulations require redundant positioning systems?
Yes. ISO 26262 and ISO 21448 require sufficient redundancy to cover sensor faults, and regulations like GB 44721-2026 and UN Regulation No. 157 mandate backup positioning and fail-safe behavior for higher automation levels.
Is GNSS jamming a realistic concern for commercial autonomous fleets?
It is a design consideration serious enough that anti-jamming and anti-spoofing capability is increasingly built into the positioning antenna itself rather than treated as an optional accessory.
What role does satellite communication play beyond positioning?
It handles fleet-level connectivity, remote diagnostics and data offload in areas without cellular coverage, which is a separate function from moment-to-moment obstacle detection.
Why does multi-orbit terminal support matter for vehicle fleets?
A single terminal that reaches GEO, MEO and LEO, with automatic roaming to terrestrial networks, avoids locking a fleet operator into one satellite provider's coverage footprint.
About StarWin
StarWin is a Chengdu-headquartered AI-driven compound solution provider spanning communication, navigation, remote sensing and computing, built to ship as one integrated system rather than components assembled from multiple vendors. Its product range covers multi-orbit ESA and flat-panel satellite terminals, satellite IoT devices for the TianQi LEO constellation, and CRPA-based GNSS anti-jamming and anti-spoofing antennas designed to be integrated inside vehicles and terminals rather than added on afterward. StarWin's terminals and antennas have been qualified by more than a dozen satellite operators worldwide, and the company supports next-generation autonomous platform development. Roughly 40 percent of its workforce works in R&D, reflecting a focus on solving the connectivity and positioning problems underneath autonomous platforms rather than selling a single hardware component.
To learn more about how StarWin's multi-orbit terminals and anti-jamming positioning technology fit into an autonomous platform's sensor and communication stack, visit https://starwincom.com.
