Daltonomous - Cyber-Physical Defence for Autonomous Systems

Votes: 0
Views: 1216


Autonomous Systems - aircraft, unmanned aerial systems, and autonomous vehicles rely heavily on sensors (e.g., GPS, cameras, LiDAR, gyroscopes, etc.) for perception and use specialized algorithms for navigation and autonomous operations. The sensors' accuracy and reliability are crucial for these systems' safe operations. It is well documented that sensors may register anomalies or fail entirely due to electromagnetic interference, jamming, or changes in environmental factors. Electronic interference is the Achilles' heel of autonomous vehicles. This growing problem puts millions of automated systems at risk of failure.


For technology firms in the defence and transportation industries who build automated systems for land or air vehicles, we build AI-powered resiliency software that assures true sensor output for reliable navigation despite electrically-contested environments. We are solving this as a part of the $800M North American Assured Position, Navigation, and Timing market.

Our solutions use a combination of novel control theory and advanced AI techniques to detect unreliable sensors, derive appropriate safety measures to prevent any undesirable outcomes, and enable autonomous systems to safely complete their mission.

Our solutions are agnostic of the underlying sensor suite, designed for easy integration into existing navigation and feedback control modules in autonomous systems, and maintain low operational and performance overheads. We are the only solution on the market to employ three pillars of resilience: Detect, Protect, Recover.


Our Vehicle Intelligent Early Warning System (VIEWS™) monitors unexpected sensor anomalies caused due to environmental changes, sensor malfunction, or adversarial activities. VIEWS uses model-based methods to estimate true values and compare them with observed values to detect sensor anomalies in real time. The key advantage of VIEWS is its ability to differentiate between sensor anomalies and natural fluctuations, significantly reducing false positives. VIEWS employs AI powered root cause analysis to identify the source of anomalies ensuring precise diagnostics and records the anomalies for post-mission analysis.


Our anomaly response system (ARS™) identifies and isolates the affected sensors by analyzing the outputs of VIEWS. ARS then reconstructs the position of the autonomous system using the remaining reliable sensors. ARS uses an advanced ML-enabled sensor fusion method to calculate the measurements of the missing or faulty sensor. For example, if GPS is compromised, our recovery system isolates GPS and uses the remaining reliable (e.g., camera, LiDAR, IMU) for accurate navigation.


Based on the outputs of VIEWS and ARS, our Autonomous Recovery Control (ARC™) isolates the compromised sensors and uses the remaining reliable sensors to navigate and complete the mission.

Our recovery strategy includes the following steps:

  1. Trustworthy Historical Data: Our recovery system records historical data to estimate future positions of the system.
  2. AI-Powered State Estimation: Employ a robust time-series modeling technique to estimate future positions of the system. Our time-series modeling approach derives accurate state estimation even with reduced low-dimensional sensor signals.
  3. Context-Aware Decision Making: Apply a combination of control theory and reinforcement learning to understand the context of operation and make safe navigation decisions.



Learn how to vote for your favorites.


  • Name:
    Alex Dalton
  • Type of entry:
    Team members:
    • Pritam Dash
    • Prasanna Karthik
  • Patent status: