v2.2.2

感知

• 新的车道线检测模型，增加感知区域。
• 修改感知bug

功能

• 优化自动变道辅助，解决连续自动变道的bug。
• 放弃本地文件通讯，使用共享内存进行通信，提高通讯速度、效率
• 增加对本地硬件设备的监控
• 解决进程意外闪退的问题
• 重新加入v1.x的导航地图循迹功能

其它

• 增加日志文件

Perception

• Implemented a new lane detection model to expand the perception area.
• Fixed perception-related bugs.

Features

• Optimized the automatic lane change assistance feature to resolve the issue of continuous automatic lane changes.
• Abandoned local file communication in favor of using shared memory for communication to enhance communication speed and efficiency.
• Added monitoring for local hardware devices.
• Addressed the issue of unexpected process crashes.
• Added the navigation map tracking feature from version 1.x.

Others

• Added log files.

Full Changelog: v2.2.1...v2.2.2

v2.2.1

功能

• 增加纵向规划
• 纵向控制PID调优

其它

• 重构软件框架

Features

• Add longitudinal planning
• Fine-tune longitudinal control PID

Others

• Refactor software framework

Full Changelog: v2.2.0...v2.2.1

v2.2.0

感知

• 新的车道线检测模型
• 新增目标检测算法yolov8，检测类别包括轿车、公交车、卡车、红绿灯

功能

• 新增前向碰撞预警
• 修改自动变道辅助的bug

其它

增加物体3D可视化

Perception

• New lane detection model.
• Added object detection algorithm YOLOv8, which detects categories including cars, buses, trucks, and traffic lights.

Features

• Added forward collision warning.
• Fixed a bug in the lane change assist feature.

Others

Added 3D object visualization.

Full Changelog: v2.1.1...v2.2.0

v2.1.1

感知

• 车道线检查权重沿用v2.1.0

功能

• 修复自动变道辅助第二阶段轨迹激变
• 车道居中控制及自动变道辅助限速提高到80km/h

其它

• 更新用户界面

Perception

• Lane Line Inspection Weight remains the same as v2.1.0.

Features

• Fixed abrupt trajectory changes in the second phase of automatic lane change assistance.
• Lane centering control and automatic lane change assistance speed limit increased to 80 km/h.

Others

• Updated user interface.

Full Changelog: v2.1.0...v2.1.1

Full Changelog: v2.1.0...v2.1.1

v2.1.0

感知

• 训练泛化性能更强的模型，无车道线时可将路沿识别为车道线，并且夜间无车灯情况下仍能识别车道线
• 去除在v1.x版本中使用的paddleocr，使用Telemetry插件对车速及其它车况信息进行收集

功能

• 进一步完善变道辅助，进行更平滑的变道控制，支持低于70km/s的变道辅助
• 提高用户的控制权限，当检查到用户操纵方向盘和油门时，自动退出自动驾驶模式

其它

• 去除对pytorch的依赖

Perception

• Train models with stronger generalization performance, capable of recognizing roadside edges as lane lines when there are no lane markings, and still able to identify lane lines in low-light conditions without headlights.
• Remove the use of PaddleOCR, which was used in version v1.x, and instead, use the Telemetry plugin to collect data on vehicle speed and other vehicle conditions.

Features

• Further enhance lane-changing assistance, allowing lane changes at speeds below 70 km/h.
• Increase user control authority, automatically exiting autonomous driving mode when the system detects user manipulation of the steering wheel and throttle.

Others

• Eliminate the dependency on PyTorch.

What's Changed

• Fix indexes for some files by @eiyooooo in #59

Full Changelog: v2.0.0...v2.1.0

v2.0.0

感知

• 采用bev下的车道线识别算法
• 改用onnxruntime进行推理
• 感知响应时间缩短，GTX1650上推理时间低于0.07s

功能

• 实现了低于75km/h的车道保持辅助（LKA）
• 实现了低于50km/h的变道辅助（LCA）

其它

• 构建用户UI
• 制作logo
• 采用多进程结构，降低代码耦合

Perception

• Adopted lane detection algorithm under BEV (Bird's Eye View)
• Switched to ONNX Runtime for inference
• Reduced perception response time, inference time on GTX 1650 is below 0.07s

Features

• Implemented lane-keeping assistance (LKA) for speeds below 75 km/h
• Implemented lane change assistance (LCA) for speeds below 50 km/h

Others

• Built user UI (User Interface)
• Created logo
• Utilized a multi-process structure to decrease code coupling

v1.1

更新内容：

1. 更多的并行策略以增加推理速度
2. 可以使用车辆前向物体检测线（FCW）代替CLRNET车道线检测
3. 增加车道场景及天气感知
4. 更多的交互操作
5. 手动变道
6. 删除深度估计及场景分割接口

车道线检测和目标检测权重和v1.0相同。

v1.0

Add files via upload