Search

With the rapid development of autonomous driving and artificial intelligence technology, end-to-end autonomous driving technology has become a research hotspot. This thesis aims to explore the application of deep reinforcement learning in the realizing of end-to-end autonomous driving. We built a deep reinforcement learning virtual environment in the Carla simulator, and based on it, we trained a policy model to control a vehicle along a preplanned route. For the selection of the deep reinforcement learning algorithms, we have used the Proximal Policy Optimization algorithm due to its stable performance. Considering the complexity of end-to-end autonomous driving, we have also carefully designed a comprehensive reward function to train the policy model more efficiently. The model inputs for this study are of two types: firstly, real-time road information and vehicle state data obtained from the Carla simulator, and secondly, real-time images captured by the vehicle's front camera. In order to understand the influence of different input information on the training effect and model performance, we conducted a detailed comparative analysis. The test results showed that the accuracy and significance of the information has a significant impact on the learning effect of the agent, which in turn has a direct impact on the performance of the model. Through this study, we have not only confirmed the potential of deep reinforcement learning in the field of end-to-end autonomous driving, but also provided an important reference for future research and development of related technologies.

With the rapid development of autonomous driving and artificial intelligence technology, end-to-end autonomous driving technology has become a research hotspot. This thesis aims to explore the application of deep reinforcement learning in the realizing of end-to-end autonomous driving. We built a deep reinforcement learning virtual environment in the Carla simulator, and based on it, we trained a policy model to control a vehicle along a preplanned route. For the selection of the deep reinforcement learning algorithms, we have used the Proximal Policy Optimization algorithm due to its stable performance. Considering the complexity of end-to-end autonomous driving, we have also carefully designed a comprehensive reward function to train the policy model more efficiently. The model inputs for this study are of two types: firstly, real-time road information and vehicle state data obtained from the Carla simulator, and secondly, real-time images captured by the vehicle's front camera. In order to understand the influence of different input information on the training effect and model performance, we conducted a detailed comparative analysis. The test results showed that the accuracy and significance of the information has a significant impact on the learning effect of the agent, which in turn has a direct impact on the performance of the model. Through this study, we have not only confirmed the potential of deep reinforcement learning in the field of end-to-end autonomous driving, but also provided an important reference for future research and development of related technologies.

The thesis is a study of how architecture brings the fantasy from drawings to the reality as a building. A library is a good choice for me because of the most familiar experience that accompanied with me so far is learning. The key of a library, in this case, should be "sharing the knowledge", it will from book, discovery and discussion.

Zhuanglei Gao,1 Zhaoxia Li,2 Jieke Yan,3 Peilin Wang1 1Department of General Surgery, 2Department of Pediatrics, 3Department of Renal Transplantation, The Second Hospital of Shandong University, Jinan, Shandong, People’s Republic of China Abstract: For targeted gastric carcinoma therapy, hyaluronic acid (HA)-modified layer-by-layer nanoparticles (NPs) are applied for improving anticancer treatment efficacy and reducing toxicity and side effects. The aim of this study was to develop HA-modified NPs for the co-loading of irinotecan (IRN) and 5-fluorouracil (5-FU). A novel polymer–chitosan (CH)–HA hybrid formulation (HA–CH–IRN/5-FU NPs) consisting of poly(D,L-lactide-co-glycolide) (PLGA) and IRN as the core, CH and 5-FU as a shell on the core and HA as the outmost layer was prepared. Its morphology, average size, zeta potential and drug encapsulation ability were evaluated. Human gastric carcinoma cells (MGC803 cells) and cancer-bearing mice were used for the testing of in vitro cytotoxicity and in vivo antitumor efficiency of NPs. HA–CH–IRN/5-FU NPs displayed enhanced antitumor activity in vitro and in vivo than non-modified NPs, single drug-loaded NPs and drugs solutions. The results demonstrate that HA–CH–IRN/5-FU NPs can achieve impressive antitumor activity and the novel targeted drug delivery system offers a promising strategy for the treatment of gastric cancer. Keywords: gastric carcinoma, irinotecan, 5-fluorouracil, hyaluronic acid, layer-by-layer nanoparticles

Regulation of gene expression can occur at the pre-mRNA level by alternative splicing. As a type of alternative splicing, intron retention received attention only in recent years. Specifically, there are few reports on endogenous terminal intron retention event in different gene transcripts and only one report of antisense oligonucleotide induced retention of the terminal intron in SMN transcripts. So far, there is no report of induced terminal intron retention other gene transcripts. Terminal intron retention is proposed to be inducible in human gene transcripts using antisense oligonucleotides, with an effect on expression at the mRNA and/or protein levels. Five gene transcripts, LMNA, LMNC, ITGA4, SOD1, and DMD, were selected and for each transcript, the terminal intron/exon splice site and exon-internal sequences of the last exon were targeted for antisense oligonucleotide annealing. It was found that terminal intron retention is inducible in LMNC and ITGA4 transcripts, and possibly LMNA transcripts. Inclusion of the terminal intron decreases full-length LMNC and ITGA4 mRNA expression whereas its effect on protein expression is undetermined. Using the PMO chemistry resulted in a reduced effect on terminal intron retention compared to when 2’-OMeAOs were used. In addition, in both LMNC and ITGA4 transcripts, terminal intron retention led to approximately 3% to 13% decrease in full-length mRNA expression. Changing transfection parameters such as antisense oligonucleotide concentrations, transfection reagent, and transfection duration can also influence transfection outcome. Possible features that may predict terminal intron retention, including but not limited to, splicing motifs, intron length and splice site strengths, were compared between transcripts with and without retained terminal intron. However, no specific features for retention of the terminal intron were observed. Overall, this project shows that terminal intron retention is inducible in human gene tr