Autonomous Driving & ADAS

The Indian Traffic Problem Is Not a Scaled-Down Version of Western Traffic

Autonomous driving systems trained exclusively on KITTI, nuScenes, or Waymo Open Dataset will fail spectacularly on Indian roads. This is not a data volume problem — it is a distributional shift problem. Indian traffic includes cattle and two-wheelers cutting across lanes, autorickshaws reversing into traffic, pedestrians walking against oncoming vehicles, unmarked speed breakers, and road conditions that alternate between smooth highway and pothole-ridden urban street within 500 meters. The perception and planning stack must be designed with India's scenario distribution as a first-class requirement.

Safety-critical disclaimer: ADAS and autonomous driving systems directly affect occupant and road-user safety. All production deployments must pass ARAI/ICAT homologation, and any AI-based safety system requires formal verification, fault-mode analysis (FMEA/FTA), and conformance to ISO 26262 functional safety standards. The techniques in this chapter are for engineering development; do not deploy perception or planning AI in public-road vehicles without completing the full homologation process.

Perception Pipeline Architecture

A production-grade perception stack takes raw sensor data and produces a structured world model: detected objects with class, position, velocity, and uncertainty; driveable surface estimates; and traffic sign/signal states.

Sensor Modalities

For Indian conditions, the camera + radar combination (no LiDAR) is the pragmatic entry point for Level 2+ ADAS — it matches the cost envelope of vehicles in the ₹8–15 lakh segment (Tata Nexon, Mahindra Scorpio N, Maruti Fronx territory). LiDAR is reserved for L3/L4 programs and robotaxi development.

Object Detection: Camera Branch

The backbone of camera-based detection for automotive is BEV (Bird's Eye View) perception, where multiple camera views are transformed into a unified top-down representation. Key architectures:

For Indian OEM programs, the practical constraint is the ADAS ECU compute budget — Mobileye EyeQ5 / NVIDIA Orin class processors. BEVDepth variants with INT8 quantization fit within the 20–40 TOPS envelope available.

# Inference pipeline structure for a typical camera-radar ADAS stack
# Open data/adas-perception-logs.json for sample Indian traffic scenarios

import json

with open("data/adas-perception-logs.json") as f:
    scenario = json.load(f)["scenarios"][0]  # "Pune ring road, 2-wheeler cut-in"

# Each scenario has:
# - camera_frames: list of image paths
# - radar_tracks: [{id, range_m, azimuth_deg, velocity_mps, rcs_dbsm}]
# - ground_truth: [{object_class, position_xyz, velocity_xyz, bbox_3d}]
# - scenario_tags: ["two_wheeler", "cut_in", "urban", "peak_hour"]

print(f"Scenario: {scenario['description']}")
print(f"Objects: {len(scenario['ground_truth'])}")
print(f"Tags: {scenario['scenario_tags']}")

Sensor Fusion Architectures

Sensor fusion combines measurements from multiple sensors to produce estimates more accurate and robust than any single sensor alone. There are three architectural patterns:

Early Fusion (Raw Data Fusion)

Concatenate raw or near-raw sensor data before the detection network. Camera feature maps and LiDAR point clouds are projected into a common BEV representation and processed jointly. Best accuracy, highest compute, hardest to debug.

Mid Fusion (Feature Fusion)

Each sensor branch extracts features independently; features are fused at an intermediate layer. More modular — camera branch can be updated without retraining LiDAR branch. Current industry standard for camera+LiDAR systems.

Late Fusion (Track Fusion)

Each sensor branch produces tracked objects independently; an association algorithm merges the track lists. Most interpretable, easiest to validate for functional safety (each branch testable independently), but misses cross-modal correlations.

For ISO 26262 ASIL-B/D compliance, late fusion with redundant independent channels is preferred — each channel can be individually certified, and the fusion layer implements a voter/monitor architecture.

Kalman Filter vs Neural Track Fusion

The Extended Kalman Filter (EKF) has been the workhorse of sensor fusion for 30 years. Neural approaches (Transformer-based tracking like MUTR3D, MOTR) show better performance on complex scenarios but are harder to bound formally:

For ADAS L2 production programs at Indian OEMs, EKF with radar-primary tracking and camera-based object class enrichment is the validated path. Neural fusion is appropriate for L4 research programs.

Path Planning for Indian Traffic

The standard Frenet-frame path planning (sampling-based or optimization-based lateral/longitudinal profiles) works well for structured highway scenarios. Indian urban scenarios require extensions:

Unprotected Intersections Without Traffic Signals

A significant fraction of urban Indian intersections have no signals and rely on implicit negotiation. A rule-based planner will either deadlock (perpetually yielding) or proceed aggressively. Learned planners trained on Indian intersection data handle the implicit negotiation better:

Prompt: "I am designing a path planner for unprotected T-intersections in Bengaluru.
Common scenarios: autorickshaw moving slowly into intersection from left, motorcycle
overtaking from blind spot on right, pedestrian crossing mid-intersection.

Design the state machine for the intersection approach:
- Define the states, transitions, and timeout conditions
- Specify what each sensor input maps to state transitions
- Identify the minimum gap acceptance criteria for each opposing traffic type
- Flag which conditions require mandatory stop vs. creep-and-assess"

Speed Breaker Detection

India's unmarked speed breakers are a Level 2 ADAS-specific challenge. Radar does not detect them. LiDAR detects them at close range only. Camera-based detection with depth estimation is the primary sensor. Key ML task: binary classification + height estimation from monocular camera images, trained on Indian road imagery.

ARAI and ICAT Homologation

In India, ADAS systems require type approval from ARAI (Automotive Research Association of India, Pune) or ICAT (International Centre for Automotive Technology, Manesar). The relevant AIS standards:

CMVR (Central Motor Vehicles Rules) mandates AEB for M1 category vehicles above 3.5T GVW from April 2023, and is expanding. AIS-171 test scenarios use Euro NCAP-aligned test protocols but add specific Indian test conditions (unpaved road surfaces, 45°C ambient temperature).

V2X: Vehicle-to-Everything Communication

V2X (Vehicle-to-Everything) is the communication infrastructure that allows vehicles to share position, speed, and intent data with each other (V2V), with infrastructure (V2I), and with pedestrians (V2P). India's approach:

For ADAS engineers, V2X data augments onboard perception: an intersection controller can broadcast signal phase and timing (SPAT messages) 300m before the intersection, enabling predictive deceleration that pure camera systems cannot achieve.

Key Takeaways