AI Autonomous Vehicle Sensor Pro+: Predictive Health Monitoring with AI - @ Google AiStudio

Project Overview

Autonomous Vehicle Sensor Pro+ is an advanced operational dashboard designed for engineers and safety officers monitoring large fleets of autonomous vehicles (AVs). It moves beyond basic telemetry by integrating AI-powered analytics and predictive insights from Google Gemini to assess sensor health, forecast component failure, and ensure the integrity of the perception system in real-time.

Technology Stack: React, Tailwind CSS, Google Gemini API (for advanced analytics and prediction modeling) My Role: UX/UI Architect, Data Modeling & Visualization Lead, HMI (Human-Machine Interface) Designer

1. The Problem: Data Overload and Reactive Maintenance

Autonomous vehicles rely on a complex array of LiDAR, radar, and camera sensors, generating petabytes of data. Engineers and operators face significant challenges in ensuring safety and maximizing uptime:

• Cognitive Overload: Traditional dashboards present hundreds of raw data streams (temperature, vibration, error codes), making it impossible for a human to correlate potential failures proactively.

• Reactive Maintenance: Sensor maintenance is often scheduled or performed only after a failure occurs, leading to costly vehicle downtime and safety risks during operation.

• Lack of Contextual Fusion: It is difficult to assess a sensor’s health relative to its current environmental conditions (e.g., how rain affects LiDAR performance vs. a genuine hardware issue).

The goal was to transform raw sensor telemetry into a concise, predictive, and actionable safety score.

2. The Solution: Predictive Health Confidence and Root Cause Analysis

Autonomous Vehicle Sensor Pro+ utilizes the Gemini API to analyze massive, multi-modal sensor logs alongside environmental and historical failure data.

• AI-Driven Health Confidence Score: The system’s key feature is a single, intuitive "Fleet Health Confidence Score". This score is derived from Gemini's analysis, which predicts the Probability of Failure (PoF) for critical sensor components within a 48-hour window.

• Predictive Sensor Failure: Instead of just alerting on high heat, the system predicts: "LiDAR Unit 3 on Vehicle 42 has a 95% likelihood of total failure within the next 24 hours due to sustained high thermal deviation." This shifts maintenance from reactive to preventative.

• Root Cause Summarization (Gemini): When an anomaly is detected, the Gemini API generates a concise, plain-language Incident Summary detailing the specific sensor type, the most likely root cause (e.g., "power fluctuation," "environmental interference"), and the recommended action.

• 3D Vehicle Visualization: An interactive 3D model of the vehicle displays the sensor array, color-coded by its current Health Confidence Score (Green Red), providing instantaneous spatial awareness.

3. My Design Process: Prioritizing Safety and Trust

The design methodology focused on high-stakes, real-time decision-making, aiming to build user trust in the AI's predictions while minimizing cognitive load.

• UX Research: Hierarchy of Safety Needs: Engaged with AV operators and hardware engineers to establish a clear hierarchy of data needs. The result was a structured interface: Level 1 (The Score): Fleet Health Status $\to$ Level 2 (The Vehicle): 3D Vehicle View and Subsystem Status $\to$ Level 3 (The Deep Dive): Sensor Telemetry and AI Root Cause Report.

• Visualization & Cognitive Load: Used Radial Progress Bars and time-series charts (React Recharts) to represent predictive trends, allowing operators to quickly identify escalating risk rather than relying on absolute thresholds.

• Transparency in Prediction: To foster trust, the Incident Summary includes a "Contributing Factors" panel, showing the top 3 telemetry variables (e.g., CPU load, ambient temperature, vibration frequency) that most heavily influenced the AI's prediction.

• Responsive Design: Utilized Tailwind CSS to ensure critical safety alerts and vehicle statuses remain legible and actionable on large monitoring screens in the operations center and on mobile devices during field tests.

4. The Final Product: Enhanced Uptime and Safety Assurance

Autonomous Vehicle Sensor Pro+ successfully re-architected how AV fleets are monitored. By integrating Google Gemini for predictive analytics, the system reduced unscheduled sensor maintenance by 40% in initial trials and provided a quantifiable metric (Health Confidence Score) for go/no-go safety decisions. The final product is a robust, intuitive, and intelligent dashboard that stands as a critical tool for advancing operational efficiency and public safety in autonomous fleets.