• Journal of information and communication convergence engineering
• /
• v.21 no.4
• /
• pp.346-350
• /
• 2023

Trajectory prediction is an essential element for driving autonomous vehicles, and various trajectory prediction models have emerged with the development of deep learning technology. Convolutional neural network (CNN) is the most commonly used neural network architecture for extracting the features of visual images, and the latest models exhibit high performances. This study was conducted to identify an efficient CNN backbone model among the components of deep learning models for trajectory prediction. We changed the existing CNN backbone network of multiple-trajectory prediction models used as feature extractors to various state-of-the-art CNN models. The experiment was conducted using nuScenes, which is a dataset used for the development of autonomous vehicles. The results of each model were compared using frequently used evaluation metrics for trajectory prediction. Analyzing the impact of the backbone can improve the performance of the trajectory prediction task. Investigating the influence of the backbone on multiple deep learning models can be a future challenge.

• Proceedings of the KIEE Conference
• /
• 2003.11b
• /
• pp.123-126
• /
• 2003

This paper presents an speed reference trajectory planning methods for vibration suppression in a t-mass resonant system which has a flexible coupling between a load and a driving motor. Due to this flexibility, the system often suffers vibration especially when the motor is controlled for higher speed command. The steady state conditions are utilized to derive desired load speed trajectory which does not cause the torsional vibration. And the desired motor speed trajectory is synthesized base on the relation between load and motor speed. The simulation and experiment result suggest that the proposed method effectively suppress the vibration.

• Journal of Institute of Control, Robotics and Systems
• /
• v.10 no.8
• /
• pp.696-702
• /
• 2004

An optimal trajectory generation algorithm for capturing a moving object by a mobile robot in real-time is proposed in this paper. The linear and rotational velocities of the moving object are estimated using the Kalman filter, as a state estimator. For the estimation, the moving object is tracked by a 2-DOF active camera mounted on the mobile robot, which enables a mobile manipulator to track the mobile robot until the capturing moment. The optimal trajectory for capturing the moving object is dependent on the initial conditions of the mobile robot as well as the moving object. Therefore, real-time trajectory planning for the mobile robot is definitely required for the successful capturing of the moving object. The performance of proposed algorithm is verified through the real experiments and the superiority is demonstrated by comparing to other algorithms.

• Journal of the Korean Society for Precision Engineering
• /
• v.23 no.7 s.184
• /
• pp.84-92
• /
• 2006

This paper deals with determination of motions of a humanoid robot using genetic algorithm. A humanoid robot has some problems of the structural instability basically. So, we have to consider the stable walking gait in gait planning. Besides, it is important to make the smoothly optimal gait for saving the electric power. A mobile robot has battery to move autonomously. But a humanoid robot needs more electric power in order to drive many joints. So, if movements of walking joint don't maintain optimally, it is hard to sustain the battery power during the working period. Also, if a gait trajectory doesn't have optimal state, the expected lift span of joints tends to be decreased. Also, if a gait trajectory doesn't have optimal state, the expected lift span of joints tends to be decreased. To solve these problems, the genetic algorithm is employed to guarantee the optimal gait trajectory. The fitness functions in a genetic algorithm are introduced to find out optimal trajectory, which enables the robot to have the less reduced jerk of joints and get smooth movement. With these all process accomplished by PC-based program, the optimal solution could be obtained from the simulation. In addition, we discuss the design consideration fur the joint motion and distributed computation of tile humanoid, ISHURO, and suggest its result such as structure of the network and a disturbance observer.

• Journal of Environmental Science International
• /
• v.11 no.7
• /
• pp.669-677
• /
• 2002

This study was conducted to investigate the characteristics of role of transport on aerosol concentration at Crate Lake, Oregon USA for 1988.3~1999. 5. The IMPROVE program is a cooperative measurement effort governed by a steering committee composed of representatives from USA federal and regional-state organizations. Also IMPROVE sampler is designed to obtain a complete signature of the composition of the airborne particles affecting visibility. According to 10-day backward isentropic trajectory analysis, the frequency of local, marine and Asian trajectory showed 33.1%(335 cases), 47%(478 cases), 5.2%(53 cases) respectively. The monthly variation of nss $SO_4^{2-}$, nss S, $NO_3^-$, K and C showed the double peak pattern, high in April~May and August~september and showed the lowest concentration in Winter. The other constituents concentration except for Cl$^{[-10]}$ , Na, Mg was high in local trajectory than marine trajectory. A ratio nss $SO_4^{2-}$ to $SO_4^{2-}$ was 90.5% in marine trajectory and 98% in local trajectory. It suggest that the aerosol in Crater Lake was effected by salt. The annual mean concentration of nss $SO_4^{2-}$ and nss S decreased but the springtime concentration increased.

• Journal of Power Electronics
• /
• v.13 no.1
• /
• pp.51-58
• /
• 2013

This paper proposes the power conversion mechanism of a bailer-charge-transfer zero-current-switching (CT-ZCS) circuit. The operation modes are analyzed and researched using state trajectory equations. The topology of CT-ZCS based on soft-switching inverters offers some merits such as: tracking the input reference signal dynamically, bearing load shock and short circuit, multiplying inverter N+1 redundancy parallel, coordinating power balance for easy control, and soft-switching commutation for high efficiency and large capacity. These advantages are distinctive from conventional inverter topologies and are especially demanded in AC drives: new energy generation and grid, distributed generation systems, switching power amplifier, active power filter, and reactive power compensation and so on. Prototype is manufactured and experiment results show the feasibility and dynamic voltage-tracking characteristics of the topology.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.2033-2036
• /
• 2004

A novel approach to estimate the real time moving trajectory of an object is proposed in this paper. The object position is obtained from the image data of a CCD camera, while a state estimator predicts the linear and angular velocities of the moving object. To overcome the uncertainties and noises residing in the input data, a Kalman filter and neural networks are utilized. Since the Kalman filter needs to approximate a non-linear system into a linear model to estimate the states, there always exist errors as well as uncertainties again. To resolve this problem, the neural networks are adopted in this approach, which have high adaptability with the memory of the input-output relationship. Kohonen Network(Self-Organized Map) is selected to learn the motion trajectory since it is spatially oriented. The superiority of the proposed algorithm is demonstrated through the real experiments.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.1587-1592
• /
• 2004

In the area of tracking control, it is important to design not only the controllers but also the trajectories to which a system has to follow. Position in the form of the $5^{th}$ order polynomial is often used with constraints of initial and final states. Smooth ending with possible minimum time is important for many systems to be away from vibrations or jerky motions. A simple polynomial-like trajectory generation method based on zero final state constraints is suggested and named ZSPOT. The effects of suggested method are shown through experiments in which a system follows an easy and computationally light reference trajectory.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.1190-1195
• /
• 2003

In this paper, it is shown how a robot with n passive trailers can be controlled in backward direction. When driving backward direction, a kinematic model of the system is represented highly nonlinear equations. The problem is formulated as a trajectory following problem, rather than control of independent generalized coordinates. Also, the state and input saturation problems are formulated as a trajectory generation problem. The trajectory is traced by a rear hinge point of the last trailer, and reference trajectories include line segments, circular shapes and rectangular turns. Experimental verifications were carried out with the PSR-2(public service robot $2^{nd}$ version) with three passive trailers. Experimental result showed that the backward motion control can be successfully carried out using the proposed control scheme.

• Information and Communications Magazine
• /
• v.29 no.8
• /
• pp.72-84
• /
• 2012

This paper introduces three vehicle trajectory-based data forwarding schemes, tailored for vehicular ad hoc networks. Nowadays GPS-based navigation systems are popularly used for providing efficient driving paths for drivers. With the driving paths called vehicle trajectories, we can make data forwarding schemes more efficient, considering the micro-scoped mobility of individual vehicles in road networks as well as the macro-scoped mobility of vehicular traffic statistics. This paper shows why the vehicle trajectory is a key ingredient in the design of the vehicle-to-infrastructure, infrastructure-to-vehicle, and vehicle-to-vehicle data forwarding schemes over multihop. Through the mathematical formulation, the key design techniques are shown for three forwarding schemes based on vehicle trajectory, compared with a state-of- the- art data forwarding scheme based on only vehicular traffic statistics.